diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/README.md b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..45089d5298dfb507172e185232886d62c3f35162
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/README.md
@@ -0,0 +1,58 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: transformers
+model_name: Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1
+tags:
+- generated_from_trainer
+- sft
+- trl
+licence: license
+---
+
+# Model Card for Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1
+
+This model is a fine-tuned version of [Qwen/Qwen3.5-2B-Base](https://huggingface.co/Qwen/Qwen3.5-2B-Base).
+It has been trained using [TRL](https://github.com/huggingface/trl).
+
+## Quick start
+
+```python
+from transformers import pipeline
+
+question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
+generator = pipeline("text-generation", model="None", device="cuda")
+output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
+print(output["generated_text"])
+```
+
+## Training procedure
+
+[
](https://wandb.ai/katriin-kukk/Cross_lingual_morphological_generalization/runs/0j86h051)
+
+
+
+This model was trained with SFT.
+
+### Framework versions
+
+- TRL: 0.29.0
+- Transformers: 5.5.4
+- Pytorch: 2.10.0
+- Datasets: 4.6.1
+- Tokenizers: 0.22.2
+
+## Citations
+
+
+
+Cite TRL as:
+
+```bibtex
+@software{vonwerra2020trl,
+ title = {{TRL: Transformers Reinforcement Learning}},
+ author = {von Werra, Leandro and Belkada, Younes and Tunstall, Lewis and Beeching, Edward and Thrush, Tristan and Lambert, Nathan and Huang, Shengyi and Rasul, Kashif and Gallouédec, Quentin},
+ license = {Apache-2.0},
+ url = {https://github.com/huggingface/trl},
+ year = {2020}
+}
+```
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1284/README.md b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1284/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1284/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1284/adapter_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1284/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c2a7efe7af0ceadd86d0f851dc87d465191c777
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1284/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 64,
+ "lora_bias": false,
+ "lora_dropout": 0.04180832058159435,
+ "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": 64,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "v_proj",
+ "o_proj",
+ "q_proj",
+ "gate_proj",
+ "k_proj",
+ "down_proj",
+ "up_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1284/chat_template.jinja b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1284/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1284/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1284/tokenizer_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1284/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1284/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1284/trainer_state.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1284/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6145e543cda7a2ee7877fd8ceb320723020255f3
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1284/trainer_state.json
@@ -0,0 +1,317 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 3.0,
+ "eval_steps": 500,
+ "global_step": 1284,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 1.8341716039180755,
+ "epoch": 0.11689070718877849,
+ "grad_norm": 1.1992322206497192,
+ "learning_rate": 5.366728168873013e-05,
+ "loss": 1.6918942260742187,
+ "mean_token_accuracy": 0.631798365265131,
+ "num_tokens": 153689.0,
+ "step": 50
+ },
+ {
+ "entropy": 0.9513179230690002,
+ "epoch": 0.23378141437755698,
+ "grad_norm": 1.1290454864501953,
+ "learning_rate": 0.00010842981402416902,
+ "loss": 0.8867578887939453,
+ "mean_token_accuracy": 0.7587438315153122,
+ "num_tokens": 310548.0,
+ "step": 100
+ },
+ {
+ "entropy": 0.8510265004634857,
+ "epoch": 0.3506721215663355,
+ "grad_norm": 0.7732136845588684,
+ "learning_rate": 0.00016319234635960792,
+ "loss": 0.7856448364257812,
+ "mean_token_accuracy": 0.7769897204637527,
+ "num_tokens": 468560.0,
+ "step": 150
+ },
+ {
+ "entropy": 0.8142367601394653,
+ "epoch": 0.46756282875511396,
+ "grad_norm": 0.7580806016921997,
+ "learning_rate": 0.00021795487869504684,
+ "loss": 0.7461666870117187,
+ "mean_token_accuracy": 0.7888966089487076,
+ "num_tokens": 619552.0,
+ "step": 200
+ },
+ {
+ "entropy": 0.772578022480011,
+ "epoch": 0.5844535359438925,
+ "grad_norm": 0.5330358147621155,
+ "learning_rate": 0.00027271741103048575,
+ "loss": 0.7142900848388671,
+ "mean_token_accuracy": 0.7959607627987861,
+ "num_tokens": 771972.0,
+ "step": 250
+ },
+ {
+ "entropy": 0.7481484657526016,
+ "epoch": 0.701344243132671,
+ "grad_norm": 0.8242517709732056,
+ "learning_rate": 0.00032747994336592464,
+ "loss": 0.6911422729492187,
+ "mean_token_accuracy": 0.8028205358982086,
+ "num_tokens": 928088.0,
+ "step": 300
+ },
+ {
+ "entropy": 0.741372903585434,
+ "epoch": 0.8182349503214494,
+ "grad_norm": 0.7722981572151184,
+ "learning_rate": 0.00038224247570136353,
+ "loss": 0.6908904266357422,
+ "mean_token_accuracy": 0.8018302822113037,
+ "num_tokens": 1086836.0,
+ "step": 350
+ },
+ {
+ "entropy": 0.7257227802276611,
+ "epoch": 0.9351256575102279,
+ "grad_norm": 0.7969573140144348,
+ "learning_rate": 0.0004370050080368024,
+ "loss": 0.6714310455322265,
+ "mean_token_accuracy": 0.8057891410589219,
+ "num_tokens": 1241227.0,
+ "step": 400
+ },
+ {
+ "epoch": 1.0,
+ "eval_entropy": 0.8317563653766334,
+ "eval_loss": 0.8135058283805847,
+ "eval_mean_token_accuracy": 0.7755894215850087,
+ "eval_num_tokens": 1325643.0,
+ "eval_runtime": 37.4951,
+ "eval_samples_per_second": 32.831,
+ "eval_steps_per_second": 4.107,
+ "step": 428
+ },
+ {
+ "entropy": 0.7179872094087265,
+ "epoch": 1.0514319111630626,
+ "grad_norm": 0.6224768161773682,
+ "learning_rate": 0.00046873290101955413,
+ "loss": 0.6683222961425781,
+ "mean_token_accuracy": 0.8054787133207273,
+ "num_tokens": 1393922.0,
+ "step": 450
+ },
+ {
+ "entropy": 0.6991294291615486,
+ "epoch": 1.1683226183518411,
+ "grad_norm": 0.6813111901283264,
+ "learning_rate": 0.00046837443306086947,
+ "loss": 0.6493977355957031,
+ "mean_token_accuracy": 0.8089349576830864,
+ "num_tokens": 1551442.0,
+ "step": 500
+ },
+ {
+ "entropy": 0.6951853144168854,
+ "epoch": 1.2852133255406195,
+ "grad_norm": 0.617091953754425,
+ "learning_rate": 0.0004676269147738558,
+ "loss": 0.6484123229980469,
+ "mean_token_accuracy": 0.8088837671279907,
+ "num_tokens": 1706697.0,
+ "step": 550
+ },
+ {
+ "entropy": 0.6852494943141937,
+ "epoch": 1.4021040327293979,
+ "grad_norm": 0.4677598774433136,
+ "learning_rate": 0.0004664915890374708,
+ "loss": 0.6416233062744141,
+ "mean_token_accuracy": 0.8106312158703805,
+ "num_tokens": 1865246.0,
+ "step": 600
+ },
+ {
+ "entropy": 0.6703594943881035,
+ "epoch": 1.5189947399181765,
+ "grad_norm": 0.5708025693893433,
+ "learning_rate": 0.0004649703435278991,
+ "loss": 0.6273183441162109,
+ "mean_token_accuracy": 0.8155005398392677,
+ "num_tokens": 2022060.0,
+ "step": 650
+ },
+ {
+ "entropy": 0.6676286320388317,
+ "epoch": 1.635885447106955,
+ "grad_norm": 0.5555347800254822,
+ "learning_rate": 0.00046306570757996264,
+ "loss": 0.6276700973510743,
+ "mean_token_accuracy": 0.8154828292131424,
+ "num_tokens": 2178909.0,
+ "step": 700
+ },
+ {
+ "entropy": 0.6669242936372757,
+ "epoch": 1.7527761542957334,
+ "grad_norm": 0.5260112285614014,
+ "learning_rate": 0.0004607808479816624,
+ "loss": 0.6228141784667969,
+ "mean_token_accuracy": 0.8164840793609619,
+ "num_tokens": 2329314.0,
+ "step": 750
+ },
+ {
+ "entropy": 0.6598560312390327,
+ "epoch": 1.869666861484512,
+ "grad_norm": 0.5677736401557922,
+ "learning_rate": 0.0004581195637088436,
+ "loss": 0.6214292907714843,
+ "mean_token_accuracy": 0.8171795177459716,
+ "num_tokens": 2478559.0,
+ "step": 800
+ },
+ {
+ "entropy": 0.6687870016694069,
+ "epoch": 1.9865575686732906,
+ "grad_norm": 0.6449615955352783,
+ "learning_rate": 0.00045508627960873823,
+ "loss": 0.6243909454345703,
+ "mean_token_accuracy": 0.8147780740261078,
+ "num_tokens": 2631937.0,
+ "step": 850
+ },
+ {
+ "epoch": 2.0,
+ "eval_entropy": 0.7786144471013701,
+ "eval_loss": 0.7599140405654907,
+ "eval_mean_token_accuracy": 0.7935443106409791,
+ "eval_num_tokens": 2651286.0,
+ "eval_runtime": 39.0048,
+ "eval_samples_per_second": 31.56,
+ "eval_steps_per_second": 3.948,
+ "step": 856
+ },
+ {
+ "entropy": 0.5821921329701966,
+ "epoch": 2.102863822326125,
+ "grad_norm": 0.6116960048675537,
+ "learning_rate": 0.00045168603904288863,
+ "loss": 0.5406004714965821,
+ "mean_token_accuracy": 0.8334921918921734,
+ "num_tokens": 2795200.0,
+ "step": 900
+ },
+ {
+ "entropy": 0.5937331764400006,
+ "epoch": 2.2197545295149035,
+ "grad_norm": 0.5440122485160828,
+ "learning_rate": 0.00044792449550168286,
+ "loss": 0.5515737533569336,
+ "mean_token_accuracy": 0.8314691257476806,
+ "num_tokens": 2950534.0,
+ "step": 950
+ },
+ {
+ "entropy": 0.5877273553609847,
+ "epoch": 2.3366452367036823,
+ "grad_norm": 0.6505260467529297,
+ "learning_rate": 0.0004438079032044453,
+ "loss": 0.5507744979858399,
+ "mean_token_accuracy": 0.8314063146710395,
+ "num_tokens": 3101471.0,
+ "step": 1000
+ },
+ {
+ "entropy": 0.5897100016474723,
+ "epoch": 2.4535359438924607,
+ "grad_norm": 0.54567551612854,
+ "learning_rate": 0.0004393431067007111,
+ "loss": 0.5528768157958984,
+ "mean_token_accuracy": 0.8315245220065117,
+ "num_tokens": 3258554.0,
+ "step": 1050
+ },
+ {
+ "entropy": 0.5831590622663498,
+ "epoch": 2.570426651081239,
+ "grad_norm": 0.44826796650886536,
+ "learning_rate": 0.00043453752948997376,
+ "loss": 0.5442767333984375,
+ "mean_token_accuracy": 0.8327079233527184,
+ "num_tokens": 3416088.0,
+ "step": 1100
+ },
+ {
+ "entropy": 0.5839906217157841,
+ "epoch": 2.6873173582700174,
+ "grad_norm": 0.5067696571350098,
+ "learning_rate": 0.0004293991616788285,
+ "loss": 0.5485663604736328,
+ "mean_token_accuracy": 0.8324281191825866,
+ "num_tokens": 3567251.0,
+ "step": 1150
+ },
+ {
+ "entropy": 0.5870274990797043,
+ "epoch": 2.8042080654587958,
+ "grad_norm": 0.44032806158065796,
+ "learning_rate": 0.00042393654669603217,
+ "loss": 0.5524474716186524,
+ "mean_token_accuracy": 0.8313613015413285,
+ "num_tokens": 3719464.0,
+ "step": 1200
+ },
+ {
+ "entropy": 0.5864466108381748,
+ "epoch": 2.9210987726475746,
+ "grad_norm": 0.5520864129066467,
+ "learning_rate": 0.00041815876708756964,
+ "loss": 0.5490786361694336,
+ "mean_token_accuracy": 0.8314764249324799,
+ "num_tokens": 3874423.0,
+ "step": 1250
+ },
+ {
+ "epoch": 3.0,
+ "eval_entropy": 0.6819627613990338,
+ "eval_loss": 0.7593190670013428,
+ "eval_mean_token_accuracy": 0.7857603395914102,
+ "eval_num_tokens": 3976929.0,
+ "eval_runtime": 39.3654,
+ "eval_samples_per_second": 31.271,
+ "eval_steps_per_second": 3.912,
+ "step": 1284
+ }
+ ],
+ "logging_steps": 50,
+ "max_steps": 4280,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 10,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 5.959334198185114e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1712/README.md b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1712/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1712/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1712/adapter_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1712/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c2a7efe7af0ceadd86d0f851dc87d465191c777
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1712/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 64,
+ "lora_bias": false,
+ "lora_dropout": 0.04180832058159435,
+ "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": 64,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "v_proj",
+ "o_proj",
+ "q_proj",
+ "gate_proj",
+ "k_proj",
+ "down_proj",
+ "up_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1712/chat_template.jinja b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1712/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1712/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1712/tokenizer_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1712/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1712/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1712/trainer_state.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1712/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a0ea6cfaf3f02d1143a90db14f8085903235a02
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-1712/trainer_state.json
@@ -0,0 +1,418 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 4.0,
+ "eval_steps": 500,
+ "global_step": 1712,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 1.8341716039180755,
+ "epoch": 0.11689070718877849,
+ "grad_norm": 1.1992322206497192,
+ "learning_rate": 5.366728168873013e-05,
+ "loss": 1.6918942260742187,
+ "mean_token_accuracy": 0.631798365265131,
+ "num_tokens": 153689.0,
+ "step": 50
+ },
+ {
+ "entropy": 0.9513179230690002,
+ "epoch": 0.23378141437755698,
+ "grad_norm": 1.1290454864501953,
+ "learning_rate": 0.00010842981402416902,
+ "loss": 0.8867578887939453,
+ "mean_token_accuracy": 0.7587438315153122,
+ "num_tokens": 310548.0,
+ "step": 100
+ },
+ {
+ "entropy": 0.8510265004634857,
+ "epoch": 0.3506721215663355,
+ "grad_norm": 0.7732136845588684,
+ "learning_rate": 0.00016319234635960792,
+ "loss": 0.7856448364257812,
+ "mean_token_accuracy": 0.7769897204637527,
+ "num_tokens": 468560.0,
+ "step": 150
+ },
+ {
+ "entropy": 0.8142367601394653,
+ "epoch": 0.46756282875511396,
+ "grad_norm": 0.7580806016921997,
+ "learning_rate": 0.00021795487869504684,
+ "loss": 0.7461666870117187,
+ "mean_token_accuracy": 0.7888966089487076,
+ "num_tokens": 619552.0,
+ "step": 200
+ },
+ {
+ "entropy": 0.772578022480011,
+ "epoch": 0.5844535359438925,
+ "grad_norm": 0.5330358147621155,
+ "learning_rate": 0.00027271741103048575,
+ "loss": 0.7142900848388671,
+ "mean_token_accuracy": 0.7959607627987861,
+ "num_tokens": 771972.0,
+ "step": 250
+ },
+ {
+ "entropy": 0.7481484657526016,
+ "epoch": 0.701344243132671,
+ "grad_norm": 0.8242517709732056,
+ "learning_rate": 0.00032747994336592464,
+ "loss": 0.6911422729492187,
+ "mean_token_accuracy": 0.8028205358982086,
+ "num_tokens": 928088.0,
+ "step": 300
+ },
+ {
+ "entropy": 0.741372903585434,
+ "epoch": 0.8182349503214494,
+ "grad_norm": 0.7722981572151184,
+ "learning_rate": 0.00038224247570136353,
+ "loss": 0.6908904266357422,
+ "mean_token_accuracy": 0.8018302822113037,
+ "num_tokens": 1086836.0,
+ "step": 350
+ },
+ {
+ "entropy": 0.7257227802276611,
+ "epoch": 0.9351256575102279,
+ "grad_norm": 0.7969573140144348,
+ "learning_rate": 0.0004370050080368024,
+ "loss": 0.6714310455322265,
+ "mean_token_accuracy": 0.8057891410589219,
+ "num_tokens": 1241227.0,
+ "step": 400
+ },
+ {
+ "epoch": 1.0,
+ "eval_entropy": 0.8317563653766334,
+ "eval_loss": 0.8135058283805847,
+ "eval_mean_token_accuracy": 0.7755894215850087,
+ "eval_num_tokens": 1325643.0,
+ "eval_runtime": 37.4951,
+ "eval_samples_per_second": 32.831,
+ "eval_steps_per_second": 4.107,
+ "step": 428
+ },
+ {
+ "entropy": 0.7179872094087265,
+ "epoch": 1.0514319111630626,
+ "grad_norm": 0.6224768161773682,
+ "learning_rate": 0.00046873290101955413,
+ "loss": 0.6683222961425781,
+ "mean_token_accuracy": 0.8054787133207273,
+ "num_tokens": 1393922.0,
+ "step": 450
+ },
+ {
+ "entropy": 0.6991294291615486,
+ "epoch": 1.1683226183518411,
+ "grad_norm": 0.6813111901283264,
+ "learning_rate": 0.00046837443306086947,
+ "loss": 0.6493977355957031,
+ "mean_token_accuracy": 0.8089349576830864,
+ "num_tokens": 1551442.0,
+ "step": 500
+ },
+ {
+ "entropy": 0.6951853144168854,
+ "epoch": 1.2852133255406195,
+ "grad_norm": 0.617091953754425,
+ "learning_rate": 0.0004676269147738558,
+ "loss": 0.6484123229980469,
+ "mean_token_accuracy": 0.8088837671279907,
+ "num_tokens": 1706697.0,
+ "step": 550
+ },
+ {
+ "entropy": 0.6852494943141937,
+ "epoch": 1.4021040327293979,
+ "grad_norm": 0.4677598774433136,
+ "learning_rate": 0.0004664915890374708,
+ "loss": 0.6416233062744141,
+ "mean_token_accuracy": 0.8106312158703805,
+ "num_tokens": 1865246.0,
+ "step": 600
+ },
+ {
+ "entropy": 0.6703594943881035,
+ "epoch": 1.5189947399181765,
+ "grad_norm": 0.5708025693893433,
+ "learning_rate": 0.0004649703435278991,
+ "loss": 0.6273183441162109,
+ "mean_token_accuracy": 0.8155005398392677,
+ "num_tokens": 2022060.0,
+ "step": 650
+ },
+ {
+ "entropy": 0.6676286320388317,
+ "epoch": 1.635885447106955,
+ "grad_norm": 0.5555347800254822,
+ "learning_rate": 0.00046306570757996264,
+ "loss": 0.6276700973510743,
+ "mean_token_accuracy": 0.8154828292131424,
+ "num_tokens": 2178909.0,
+ "step": 700
+ },
+ {
+ "entropy": 0.6669242936372757,
+ "epoch": 1.7527761542957334,
+ "grad_norm": 0.5260112285614014,
+ "learning_rate": 0.0004607808479816624,
+ "loss": 0.6228141784667969,
+ "mean_token_accuracy": 0.8164840793609619,
+ "num_tokens": 2329314.0,
+ "step": 750
+ },
+ {
+ "entropy": 0.6598560312390327,
+ "epoch": 1.869666861484512,
+ "grad_norm": 0.5677736401557922,
+ "learning_rate": 0.0004581195637088436,
+ "loss": 0.6214292907714843,
+ "mean_token_accuracy": 0.8171795177459716,
+ "num_tokens": 2478559.0,
+ "step": 800
+ },
+ {
+ "entropy": 0.6687870016694069,
+ "epoch": 1.9865575686732906,
+ "grad_norm": 0.6449615955352783,
+ "learning_rate": 0.00045508627960873823,
+ "loss": 0.6243909454345703,
+ "mean_token_accuracy": 0.8147780740261078,
+ "num_tokens": 2631937.0,
+ "step": 850
+ },
+ {
+ "epoch": 2.0,
+ "eval_entropy": 0.7786144471013701,
+ "eval_loss": 0.7599140405654907,
+ "eval_mean_token_accuracy": 0.7935443106409791,
+ "eval_num_tokens": 2651286.0,
+ "eval_runtime": 39.0048,
+ "eval_samples_per_second": 31.56,
+ "eval_steps_per_second": 3.948,
+ "step": 856
+ },
+ {
+ "entropy": 0.5821921329701966,
+ "epoch": 2.102863822326125,
+ "grad_norm": 0.6116960048675537,
+ "learning_rate": 0.00045168603904288863,
+ "loss": 0.5406004714965821,
+ "mean_token_accuracy": 0.8334921918921734,
+ "num_tokens": 2795200.0,
+ "step": 900
+ },
+ {
+ "entropy": 0.5937331764400006,
+ "epoch": 2.2197545295149035,
+ "grad_norm": 0.5440122485160828,
+ "learning_rate": 0.00044792449550168286,
+ "loss": 0.5515737533569336,
+ "mean_token_accuracy": 0.8314691257476806,
+ "num_tokens": 2950534.0,
+ "step": 950
+ },
+ {
+ "entropy": 0.5877273553609847,
+ "epoch": 2.3366452367036823,
+ "grad_norm": 0.6505260467529297,
+ "learning_rate": 0.0004438079032044453,
+ "loss": 0.5507744979858399,
+ "mean_token_accuracy": 0.8314063146710395,
+ "num_tokens": 3101471.0,
+ "step": 1000
+ },
+ {
+ "entropy": 0.5897100016474723,
+ "epoch": 2.4535359438924607,
+ "grad_norm": 0.54567551612854,
+ "learning_rate": 0.0004393431067007111,
+ "loss": 0.5528768157958984,
+ "mean_token_accuracy": 0.8315245220065117,
+ "num_tokens": 3258554.0,
+ "step": 1050
+ },
+ {
+ "entropy": 0.5831590622663498,
+ "epoch": 2.570426651081239,
+ "grad_norm": 0.44826796650886536,
+ "learning_rate": 0.00043453752948997376,
+ "loss": 0.5442767333984375,
+ "mean_token_accuracy": 0.8327079233527184,
+ "num_tokens": 3416088.0,
+ "step": 1100
+ },
+ {
+ "entropy": 0.5839906217157841,
+ "epoch": 2.6873173582700174,
+ "grad_norm": 0.5067696571350098,
+ "learning_rate": 0.0004293991616788285,
+ "loss": 0.5485663604736328,
+ "mean_token_accuracy": 0.8324281191825866,
+ "num_tokens": 3567251.0,
+ "step": 1150
+ },
+ {
+ "entropy": 0.5870274990797043,
+ "epoch": 2.8042080654587958,
+ "grad_norm": 0.44032806158065796,
+ "learning_rate": 0.00042393654669603217,
+ "loss": 0.5524474716186524,
+ "mean_token_accuracy": 0.8313613015413285,
+ "num_tokens": 3719464.0,
+ "step": 1200
+ },
+ {
+ "entropy": 0.5864466108381748,
+ "epoch": 2.9210987726475746,
+ "grad_norm": 0.5520864129066467,
+ "learning_rate": 0.00041815876708756964,
+ "loss": 0.5490786361694336,
+ "mean_token_accuracy": 0.8314764249324799,
+ "num_tokens": 3874423.0,
+ "step": 1250
+ },
+ {
+ "epoch": 3.0,
+ "eval_entropy": 0.6819627613990338,
+ "eval_loss": 0.7593190670013428,
+ "eval_mean_token_accuracy": 0.7857603395914102,
+ "eval_num_tokens": 3976929.0,
+ "eval_runtime": 39.3654,
+ "eval_samples_per_second": 31.271,
+ "eval_steps_per_second": 3.912,
+ "step": 1284
+ },
+ {
+ "entropy": 0.5587641523411525,
+ "epoch": 3.037405026300409,
+ "grad_norm": 0.622401773929596,
+ "learning_rate": 0.0004120754294153441,
+ "loss": 0.518932762145996,
+ "mean_token_accuracy": 0.8399260457436643,
+ "num_tokens": 4027609.0,
+ "step": 1300
+ },
+ {
+ "entropy": 0.5028628017008304,
+ "epoch": 3.1542957334891875,
+ "grad_norm": 0.577396810054779,
+ "learning_rate": 0.00040569664828459917,
+ "loss": 0.4589382171630859,
+ "mean_token_accuracy": 0.8538418188691139,
+ "num_tokens": 4180560.0,
+ "step": 1350
+ },
+ {
+ "entropy": 0.5227407096326351,
+ "epoch": 3.2711864406779663,
+ "grad_norm": 0.5113071203231812,
+ "learning_rate": 0.00039903302952663176,
+ "loss": 0.47863777160644533,
+ "mean_token_accuracy": 0.847364938557148,
+ "num_tokens": 4329954.0,
+ "step": 1400
+ },
+ {
+ "entropy": 0.5050077450275421,
+ "epoch": 3.3880771478667446,
+ "grad_norm": 0.48977184295654297,
+ "learning_rate": 0.0003920956525647558,
+ "loss": 0.46841480255126955,
+ "mean_token_accuracy": 0.8519425508379936,
+ "num_tokens": 4484354.0,
+ "step": 1450
+ },
+ {
+ "entropy": 0.5163291451334954,
+ "epoch": 3.504967855055523,
+ "grad_norm": 0.509864330291748,
+ "learning_rate": 0.000384896051992837,
+ "loss": 0.47587432861328127,
+ "mean_token_accuracy": 0.8486990982294083,
+ "num_tokens": 4638919.0,
+ "step": 1500
+ },
+ {
+ "entropy": 0.5134807989001274,
+ "epoch": 3.6218585622443014,
+ "grad_norm": 0.46544620394706726,
+ "learning_rate": 0.00037744619839702735,
+ "loss": 0.47692710876464844,
+ "mean_token_accuracy": 0.8489899519085884,
+ "num_tokens": 4795540.0,
+ "step": 1550
+ },
+ {
+ "entropy": 0.5216251534223556,
+ "epoch": 3.73874926943308,
+ "grad_norm": 0.5167004466056824,
+ "learning_rate": 0.0003697584784525874,
+ "loss": 0.4837848663330078,
+ "mean_token_accuracy": 0.847066233754158,
+ "num_tokens": 4951177.0,
+ "step": 1600
+ },
+ {
+ "entropy": 0.5254101701080799,
+ "epoch": 3.8556399766218585,
+ "grad_norm": 0.5504006743431091,
+ "learning_rate": 0.00036184567432888745,
+ "loss": 0.48759506225585936,
+ "mean_token_accuracy": 0.8457375919818878,
+ "num_tokens": 5104814.0,
+ "step": 1650
+ },
+ {
+ "entropy": 0.5107753933966159,
+ "epoch": 3.972530683810637,
+ "grad_norm": 0.421975314617157,
+ "learning_rate": 0.0003537209424368311,
+ "loss": 0.4743759536743164,
+ "mean_token_accuracy": 0.8505110186338425,
+ "num_tokens": 5265185.0,
+ "step": 1700
+ },
+ {
+ "epoch": 4.0,
+ "eval_entropy": 0.665543915002377,
+ "eval_loss": 0.7487082481384277,
+ "eval_mean_token_accuracy": 0.7999678904359991,
+ "eval_num_tokens": 5302572.0,
+ "eval_runtime": 38.5439,
+ "eval_samples_per_second": 31.938,
+ "eval_steps_per_second": 3.995,
+ "step": 1712
+ }
+ ],
+ "logging_steps": 50,
+ "max_steps": 4280,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 10,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 7.944979792148429e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2140/README.md b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2140/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2140/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2140/adapter_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2140/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c2a7efe7af0ceadd86d0f851dc87d465191c777
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2140/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 64,
+ "lora_bias": false,
+ "lora_dropout": 0.04180832058159435,
+ "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": 64,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "v_proj",
+ "o_proj",
+ "q_proj",
+ "gate_proj",
+ "k_proj",
+ "down_proj",
+ "up_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2140/chat_template.jinja b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2140/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2140/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2140/tokenizer_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2140/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2140/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2140/trainer_state.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2140/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..4a82769919aa39d3d91dee8e6dc306f0b2c9c039
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2140/trainer_state.json
@@ -0,0 +1,509 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 5.0,
+ "eval_steps": 500,
+ "global_step": 2140,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 1.8341716039180755,
+ "epoch": 0.11689070718877849,
+ "grad_norm": 1.1992322206497192,
+ "learning_rate": 5.366728168873013e-05,
+ "loss": 1.6918942260742187,
+ "mean_token_accuracy": 0.631798365265131,
+ "num_tokens": 153689.0,
+ "step": 50
+ },
+ {
+ "entropy": 0.9513179230690002,
+ "epoch": 0.23378141437755698,
+ "grad_norm": 1.1290454864501953,
+ "learning_rate": 0.00010842981402416902,
+ "loss": 0.8867578887939453,
+ "mean_token_accuracy": 0.7587438315153122,
+ "num_tokens": 310548.0,
+ "step": 100
+ },
+ {
+ "entropy": 0.8510265004634857,
+ "epoch": 0.3506721215663355,
+ "grad_norm": 0.7732136845588684,
+ "learning_rate": 0.00016319234635960792,
+ "loss": 0.7856448364257812,
+ "mean_token_accuracy": 0.7769897204637527,
+ "num_tokens": 468560.0,
+ "step": 150
+ },
+ {
+ "entropy": 0.8142367601394653,
+ "epoch": 0.46756282875511396,
+ "grad_norm": 0.7580806016921997,
+ "learning_rate": 0.00021795487869504684,
+ "loss": 0.7461666870117187,
+ "mean_token_accuracy": 0.7888966089487076,
+ "num_tokens": 619552.0,
+ "step": 200
+ },
+ {
+ "entropy": 0.772578022480011,
+ "epoch": 0.5844535359438925,
+ "grad_norm": 0.5330358147621155,
+ "learning_rate": 0.00027271741103048575,
+ "loss": 0.7142900848388671,
+ "mean_token_accuracy": 0.7959607627987861,
+ "num_tokens": 771972.0,
+ "step": 250
+ },
+ {
+ "entropy": 0.7481484657526016,
+ "epoch": 0.701344243132671,
+ "grad_norm": 0.8242517709732056,
+ "learning_rate": 0.00032747994336592464,
+ "loss": 0.6911422729492187,
+ "mean_token_accuracy": 0.8028205358982086,
+ "num_tokens": 928088.0,
+ "step": 300
+ },
+ {
+ "entropy": 0.741372903585434,
+ "epoch": 0.8182349503214494,
+ "grad_norm": 0.7722981572151184,
+ "learning_rate": 0.00038224247570136353,
+ "loss": 0.6908904266357422,
+ "mean_token_accuracy": 0.8018302822113037,
+ "num_tokens": 1086836.0,
+ "step": 350
+ },
+ {
+ "entropy": 0.7257227802276611,
+ "epoch": 0.9351256575102279,
+ "grad_norm": 0.7969573140144348,
+ "learning_rate": 0.0004370050080368024,
+ "loss": 0.6714310455322265,
+ "mean_token_accuracy": 0.8057891410589219,
+ "num_tokens": 1241227.0,
+ "step": 400
+ },
+ {
+ "epoch": 1.0,
+ "eval_entropy": 0.8317563653766334,
+ "eval_loss": 0.8135058283805847,
+ "eval_mean_token_accuracy": 0.7755894215850087,
+ "eval_num_tokens": 1325643.0,
+ "eval_runtime": 37.4951,
+ "eval_samples_per_second": 32.831,
+ "eval_steps_per_second": 4.107,
+ "step": 428
+ },
+ {
+ "entropy": 0.7179872094087265,
+ "epoch": 1.0514319111630626,
+ "grad_norm": 0.6224768161773682,
+ "learning_rate": 0.00046873290101955413,
+ "loss": 0.6683222961425781,
+ "mean_token_accuracy": 0.8054787133207273,
+ "num_tokens": 1393922.0,
+ "step": 450
+ },
+ {
+ "entropy": 0.6991294291615486,
+ "epoch": 1.1683226183518411,
+ "grad_norm": 0.6813111901283264,
+ "learning_rate": 0.00046837443306086947,
+ "loss": 0.6493977355957031,
+ "mean_token_accuracy": 0.8089349576830864,
+ "num_tokens": 1551442.0,
+ "step": 500
+ },
+ {
+ "entropy": 0.6951853144168854,
+ "epoch": 1.2852133255406195,
+ "grad_norm": 0.617091953754425,
+ "learning_rate": 0.0004676269147738558,
+ "loss": 0.6484123229980469,
+ "mean_token_accuracy": 0.8088837671279907,
+ "num_tokens": 1706697.0,
+ "step": 550
+ },
+ {
+ "entropy": 0.6852494943141937,
+ "epoch": 1.4021040327293979,
+ "grad_norm": 0.4677598774433136,
+ "learning_rate": 0.0004664915890374708,
+ "loss": 0.6416233062744141,
+ "mean_token_accuracy": 0.8106312158703805,
+ "num_tokens": 1865246.0,
+ "step": 600
+ },
+ {
+ "entropy": 0.6703594943881035,
+ "epoch": 1.5189947399181765,
+ "grad_norm": 0.5708025693893433,
+ "learning_rate": 0.0004649703435278991,
+ "loss": 0.6273183441162109,
+ "mean_token_accuracy": 0.8155005398392677,
+ "num_tokens": 2022060.0,
+ "step": 650
+ },
+ {
+ "entropy": 0.6676286320388317,
+ "epoch": 1.635885447106955,
+ "grad_norm": 0.5555347800254822,
+ "learning_rate": 0.00046306570757996264,
+ "loss": 0.6276700973510743,
+ "mean_token_accuracy": 0.8154828292131424,
+ "num_tokens": 2178909.0,
+ "step": 700
+ },
+ {
+ "entropy": 0.6669242936372757,
+ "epoch": 1.7527761542957334,
+ "grad_norm": 0.5260112285614014,
+ "learning_rate": 0.0004607808479816624,
+ "loss": 0.6228141784667969,
+ "mean_token_accuracy": 0.8164840793609619,
+ "num_tokens": 2329314.0,
+ "step": 750
+ },
+ {
+ "entropy": 0.6598560312390327,
+ "epoch": 1.869666861484512,
+ "grad_norm": 0.5677736401557922,
+ "learning_rate": 0.0004581195637088436,
+ "loss": 0.6214292907714843,
+ "mean_token_accuracy": 0.8171795177459716,
+ "num_tokens": 2478559.0,
+ "step": 800
+ },
+ {
+ "entropy": 0.6687870016694069,
+ "epoch": 1.9865575686732906,
+ "grad_norm": 0.6449615955352783,
+ "learning_rate": 0.00045508627960873823,
+ "loss": 0.6243909454345703,
+ "mean_token_accuracy": 0.8147780740261078,
+ "num_tokens": 2631937.0,
+ "step": 850
+ },
+ {
+ "epoch": 2.0,
+ "eval_entropy": 0.7786144471013701,
+ "eval_loss": 0.7599140405654907,
+ "eval_mean_token_accuracy": 0.7935443106409791,
+ "eval_num_tokens": 2651286.0,
+ "eval_runtime": 39.0048,
+ "eval_samples_per_second": 31.56,
+ "eval_steps_per_second": 3.948,
+ "step": 856
+ },
+ {
+ "entropy": 0.5821921329701966,
+ "epoch": 2.102863822326125,
+ "grad_norm": 0.6116960048675537,
+ "learning_rate": 0.00045168603904288863,
+ "loss": 0.5406004714965821,
+ "mean_token_accuracy": 0.8334921918921734,
+ "num_tokens": 2795200.0,
+ "step": 900
+ },
+ {
+ "entropy": 0.5937331764400006,
+ "epoch": 2.2197545295149035,
+ "grad_norm": 0.5440122485160828,
+ "learning_rate": 0.00044792449550168286,
+ "loss": 0.5515737533569336,
+ "mean_token_accuracy": 0.8314691257476806,
+ "num_tokens": 2950534.0,
+ "step": 950
+ },
+ {
+ "entropy": 0.5877273553609847,
+ "epoch": 2.3366452367036823,
+ "grad_norm": 0.6505260467529297,
+ "learning_rate": 0.0004438079032044453,
+ "loss": 0.5507744979858399,
+ "mean_token_accuracy": 0.8314063146710395,
+ "num_tokens": 3101471.0,
+ "step": 1000
+ },
+ {
+ "entropy": 0.5897100016474723,
+ "epoch": 2.4535359438924607,
+ "grad_norm": 0.54567551612854,
+ "learning_rate": 0.0004393431067007111,
+ "loss": 0.5528768157958984,
+ "mean_token_accuracy": 0.8315245220065117,
+ "num_tokens": 3258554.0,
+ "step": 1050
+ },
+ {
+ "entropy": 0.5831590622663498,
+ "epoch": 2.570426651081239,
+ "grad_norm": 0.44826796650886536,
+ "learning_rate": 0.00043453752948997376,
+ "loss": 0.5442767333984375,
+ "mean_token_accuracy": 0.8327079233527184,
+ "num_tokens": 3416088.0,
+ "step": 1100
+ },
+ {
+ "entropy": 0.5839906217157841,
+ "epoch": 2.6873173582700174,
+ "grad_norm": 0.5067696571350098,
+ "learning_rate": 0.0004293991616788285,
+ "loss": 0.5485663604736328,
+ "mean_token_accuracy": 0.8324281191825866,
+ "num_tokens": 3567251.0,
+ "step": 1150
+ },
+ {
+ "entropy": 0.5870274990797043,
+ "epoch": 2.8042080654587958,
+ "grad_norm": 0.44032806158065796,
+ "learning_rate": 0.00042393654669603217,
+ "loss": 0.5524474716186524,
+ "mean_token_accuracy": 0.8313613015413285,
+ "num_tokens": 3719464.0,
+ "step": 1200
+ },
+ {
+ "entropy": 0.5864466108381748,
+ "epoch": 2.9210987726475746,
+ "grad_norm": 0.5520864129066467,
+ "learning_rate": 0.00041815876708756964,
+ "loss": 0.5490786361694336,
+ "mean_token_accuracy": 0.8314764249324799,
+ "num_tokens": 3874423.0,
+ "step": 1250
+ },
+ {
+ "epoch": 3.0,
+ "eval_entropy": 0.6819627613990338,
+ "eval_loss": 0.7593190670013428,
+ "eval_mean_token_accuracy": 0.7857603395914102,
+ "eval_num_tokens": 3976929.0,
+ "eval_runtime": 39.3654,
+ "eval_samples_per_second": 31.271,
+ "eval_steps_per_second": 3.912,
+ "step": 1284
+ },
+ {
+ "entropy": 0.5587641523411525,
+ "epoch": 3.037405026300409,
+ "grad_norm": 0.622401773929596,
+ "learning_rate": 0.0004120754294153441,
+ "loss": 0.518932762145996,
+ "mean_token_accuracy": 0.8399260457436643,
+ "num_tokens": 4027609.0,
+ "step": 1300
+ },
+ {
+ "entropy": 0.5028628017008304,
+ "epoch": 3.1542957334891875,
+ "grad_norm": 0.577396810054779,
+ "learning_rate": 0.00040569664828459917,
+ "loss": 0.4589382171630859,
+ "mean_token_accuracy": 0.8538418188691139,
+ "num_tokens": 4180560.0,
+ "step": 1350
+ },
+ {
+ "entropy": 0.5227407096326351,
+ "epoch": 3.2711864406779663,
+ "grad_norm": 0.5113071203231812,
+ "learning_rate": 0.00039903302952663176,
+ "loss": 0.47863777160644533,
+ "mean_token_accuracy": 0.847364938557148,
+ "num_tokens": 4329954.0,
+ "step": 1400
+ },
+ {
+ "entropy": 0.5050077450275421,
+ "epoch": 3.3880771478667446,
+ "grad_norm": 0.48977184295654297,
+ "learning_rate": 0.0003920956525647558,
+ "loss": 0.46841480255126955,
+ "mean_token_accuracy": 0.8519425508379936,
+ "num_tokens": 4484354.0,
+ "step": 1450
+ },
+ {
+ "entropy": 0.5163291451334954,
+ "epoch": 3.504967855055523,
+ "grad_norm": 0.509864330291748,
+ "learning_rate": 0.000384896051992837,
+ "loss": 0.47587432861328127,
+ "mean_token_accuracy": 0.8486990982294083,
+ "num_tokens": 4638919.0,
+ "step": 1500
+ },
+ {
+ "entropy": 0.5134807989001274,
+ "epoch": 3.6218585622443014,
+ "grad_norm": 0.46544620394706726,
+ "learning_rate": 0.00037744619839702735,
+ "loss": 0.47692710876464844,
+ "mean_token_accuracy": 0.8489899519085884,
+ "num_tokens": 4795540.0,
+ "step": 1550
+ },
+ {
+ "entropy": 0.5216251534223556,
+ "epoch": 3.73874926943308,
+ "grad_norm": 0.5167004466056824,
+ "learning_rate": 0.0003697584784525874,
+ "loss": 0.4837848663330078,
+ "mean_token_accuracy": 0.847066233754158,
+ "num_tokens": 4951177.0,
+ "step": 1600
+ },
+ {
+ "entropy": 0.5254101701080799,
+ "epoch": 3.8556399766218585,
+ "grad_norm": 0.5504006743431091,
+ "learning_rate": 0.00036184567432888745,
+ "loss": 0.48759506225585936,
+ "mean_token_accuracy": 0.8457375919818878,
+ "num_tokens": 5104814.0,
+ "step": 1650
+ },
+ {
+ "entropy": 0.5107753933966159,
+ "epoch": 3.972530683810637,
+ "grad_norm": 0.421975314617157,
+ "learning_rate": 0.0003537209424368311,
+ "loss": 0.4743759536743164,
+ "mean_token_accuracy": 0.8505110186338425,
+ "num_tokens": 5265185.0,
+ "step": 1700
+ },
+ {
+ "epoch": 4.0,
+ "eval_entropy": 0.665543915002377,
+ "eval_loss": 0.7487082481384277,
+ "eval_mean_token_accuracy": 0.7999678904359991,
+ "eval_num_tokens": 5302572.0,
+ "eval_runtime": 38.5439,
+ "eval_samples_per_second": 31.938,
+ "eval_steps_per_second": 3.995,
+ "step": 1712
+ },
+ {
+ "entropy": 0.4414465431891494,
+ "epoch": 4.088836937463472,
+ "grad_norm": 0.5308493971824646,
+ "learning_rate": 0.0003453977915540383,
+ "loss": 0.39650299072265627,
+ "mean_token_accuracy": 0.8703725772287378,
+ "num_tokens": 5419002.0,
+ "step": 1750
+ },
+ {
+ "entropy": 0.42591195791959763,
+ "epoch": 4.20572764465225,
+ "grad_norm": 0.6111563444137573,
+ "learning_rate": 0.00033689006036415585,
+ "loss": 0.37838283538818357,
+ "mean_token_accuracy": 0.874235480427742,
+ "num_tokens": 5572969.0,
+ "step": 1800
+ },
+ {
+ "entropy": 0.44086351931095125,
+ "epoch": 4.322618351841029,
+ "grad_norm": 0.5524880290031433,
+ "learning_rate": 0.0003282118944476435,
+ "loss": 0.3920352554321289,
+ "mean_token_accuracy": 0.8694414687156677,
+ "num_tokens": 5733805.0,
+ "step": 1850
+ },
+ {
+ "entropy": 0.4380896310508251,
+ "epoch": 4.439509059029807,
+ "grad_norm": 0.5023863315582275,
+ "learning_rate": 0.0003193777227622898,
+ "loss": 0.3916081237792969,
+ "mean_token_accuracy": 0.8697139009833336,
+ "num_tokens": 5888694.0,
+ "step": 1900
+ },
+ {
+ "entropy": 0.44005885019898416,
+ "epoch": 4.556399766218585,
+ "grad_norm": 0.455997109413147,
+ "learning_rate": 0.000310402233652564,
+ "loss": 0.3955466842651367,
+ "mean_token_accuracy": 0.8693471103906631,
+ "num_tokens": 6042655.0,
+ "step": 1950
+ },
+ {
+ "entropy": 0.444269048422575,
+ "epoch": 4.673290473407365,
+ "grad_norm": 0.4542011320590973,
+ "learning_rate": 0.00030130035042769316,
+ "loss": 0.3988466262817383,
+ "mean_token_accuracy": 0.8680117425322532,
+ "num_tokens": 6203980.0,
+ "step": 2000
+ },
+ {
+ "entropy": 0.43520899042487143,
+ "epoch": 4.790181180596143,
+ "grad_norm": 0.49051445722579956,
+ "learning_rate": 0.0002920872065490688,
+ "loss": 0.39819797515869143,
+ "mean_token_accuracy": 0.8678940117359162,
+ "num_tokens": 6354604.0,
+ "step": 2050
+ },
+ {
+ "entropy": 0.429456724524498,
+ "epoch": 4.907071887784921,
+ "grad_norm": 0.5267980098724365,
+ "learning_rate": 0.0002827781204682396,
+ "loss": 0.39413917541503907,
+ "mean_token_accuracy": 0.870051506459713,
+ "num_tokens": 6507484.0,
+ "step": 2100
+ },
+ {
+ "epoch": 5.0,
+ "eval_entropy": 0.5826305385146823,
+ "eval_loss": 0.7694042921066284,
+ "eval_mean_token_accuracy": 0.7950637824349589,
+ "eval_num_tokens": 6628215.0,
+ "eval_runtime": 38.5641,
+ "eval_samples_per_second": 31.921,
+ "eval_steps_per_second": 3.993,
+ "step": 2140
+ }
+ ],
+ "logging_steps": 50,
+ "max_steps": 4280,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 10,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 9.930445154815795e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2568/README.md b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2568/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2568/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2568/adapter_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2568/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c2a7efe7af0ceadd86d0f851dc87d465191c777
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2568/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 64,
+ "lora_bias": false,
+ "lora_dropout": 0.04180832058159435,
+ "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": 64,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "v_proj",
+ "o_proj",
+ "q_proj",
+ "gate_proj",
+ "k_proj",
+ "down_proj",
+ "up_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/README.md b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..0a5b2b9ee63291ee8b2413002adf5e1b1c3636bd
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/README.md
@@ -0,0 +1,58 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: transformers
+model_name: Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2
+tags:
+- generated_from_trainer
+- sft
+- trl
+licence: license
+---
+
+# Model Card for Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2
+
+This model is a fine-tuned version of [Qwen/Qwen3.5-2B-Base](https://huggingface.co/Qwen/Qwen3.5-2B-Base).
+It has been trained using [TRL](https://github.com/huggingface/trl).
+
+## Quick start
+
+```python
+from transformers import pipeline
+
+question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
+generator = pipeline("text-generation", model="None", device="cuda")
+output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
+print(output["generated_text"])
+```
+
+## Training procedure
+
+[
](https://wandb.ai/katriin-kukk/Cross_lingual_morphological_generalization/runs/6q4yr6mt)
+
+
+
+This model was trained with SFT.
+
+### Framework versions
+
+- TRL: 0.29.0
+- Transformers: 5.5.4
+- Pytorch: 2.10.0
+- Datasets: 4.6.1
+- Tokenizers: 0.22.2
+
+## Citations
+
+
+
+Cite TRL as:
+
+```bibtex
+@software{vonwerra2020trl,
+ title = {{TRL: Transformers Reinforcement Learning}},
+ author = {von Werra, Leandro and Belkada, Younes and Tunstall, Lewis and Beeching, Edward and Thrush, Tristan and Lambert, Nathan and Huang, Shengyi and Rasul, Kashif and Gallouédec, Quentin},
+ license = {Apache-2.0},
+ url = {https://github.com/huggingface/trl},
+ year = {2020}
+}
+```
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1284/README.md b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1284/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1284/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1284/adapter_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1284/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9e45544de727016be134f25e606b7b2b87c67549
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1284/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.050044417519510255,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "v_proj",
+ "o_proj",
+ "q_proj",
+ "gate_proj",
+ "k_proj",
+ "down_proj",
+ "up_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1284/chat_template.jinja b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1284/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1284/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1284/tokenizer_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1284/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1284/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1284/trainer_state.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1284/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6afdf7c56498968853e88aaa6acf5b7760df48ab
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1284/trainer_state.json
@@ -0,0 +1,317 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 3.0,
+ "eval_steps": 500,
+ "global_step": 1284,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 1.8448095142841339,
+ "epoch": 0.11689070718877849,
+ "grad_norm": 1.4738068580627441,
+ "learning_rate": 3.0183098859435753e-05,
+ "loss": 1.6948321533203126,
+ "mean_token_accuracy": 0.6315939916670322,
+ "num_tokens": 153689.0,
+ "step": 50
+ },
+ {
+ "entropy": 0.9560942512750625,
+ "epoch": 0.23378141437755698,
+ "grad_norm": 1.378090739250183,
+ "learning_rate": 6.098217932824775e-05,
+ "loss": 0.8888477325439453,
+ "mean_token_accuracy": 0.7586692428588867,
+ "num_tokens": 310548.0,
+ "step": 100
+ },
+ {
+ "entropy": 0.848925698697567,
+ "epoch": 0.3506721215663355,
+ "grad_norm": 0.959621250629425,
+ "learning_rate": 9.178125979705974e-05,
+ "loss": 0.7850731658935547,
+ "mean_token_accuracy": 0.7774798944592476,
+ "num_tokens": 468560.0,
+ "step": 150
+ },
+ {
+ "entropy": 0.811761694252491,
+ "epoch": 0.46756282875511396,
+ "grad_norm": 0.8867514133453369,
+ "learning_rate": 0.00012258034026587175,
+ "loss": 0.7446884155273438,
+ "mean_token_accuracy": 0.7892421445250511,
+ "num_tokens": 619552.0,
+ "step": 200
+ },
+ {
+ "entropy": 0.7709847274422645,
+ "epoch": 0.5844535359438925,
+ "grad_norm": 0.5822142362594604,
+ "learning_rate": 0.00015337942073468373,
+ "loss": 0.7130480194091797,
+ "mean_token_accuracy": 0.7964023387432099,
+ "num_tokens": 771972.0,
+ "step": 250
+ },
+ {
+ "entropy": 0.7401632118225098,
+ "epoch": 0.701344243132671,
+ "grad_norm": 0.7557511925697327,
+ "learning_rate": 0.00018417850120349574,
+ "loss": 0.6857429504394531,
+ "mean_token_accuracy": 0.8041065946221352,
+ "num_tokens": 928088.0,
+ "step": 300
+ },
+ {
+ "entropy": 0.7295274111628532,
+ "epoch": 0.8182349503214494,
+ "grad_norm": 0.6769705414772034,
+ "learning_rate": 0.00021497758167230772,
+ "loss": 0.6801216125488281,
+ "mean_token_accuracy": 0.8037989619374275,
+ "num_tokens": 1086836.0,
+ "step": 350
+ },
+ {
+ "entropy": 0.710282272696495,
+ "epoch": 0.9351256575102279,
+ "grad_norm": 0.832009494304657,
+ "learning_rate": 0.00024577666214111967,
+ "loss": 0.6587177276611328,
+ "mean_token_accuracy": 0.8084024578332901,
+ "num_tokens": 1241227.0,
+ "step": 400
+ },
+ {
+ "epoch": 1.0,
+ "eval_entropy": 0.848237401672772,
+ "eval_loss": 0.7832363247871399,
+ "eval_mean_token_accuracy": 0.7817642971292719,
+ "eval_num_tokens": 1325643.0,
+ "eval_runtime": 36.8595,
+ "eval_samples_per_second": 33.397,
+ "eval_steps_per_second": 4.178,
+ "step": 428
+ },
+ {
+ "entropy": 0.7039205249230466,
+ "epoch": 1.0514319111630626,
+ "grad_norm": 0.6213153600692749,
+ "learning_rate": 0.00026362079548206905,
+ "loss": 0.6527813720703125,
+ "mean_token_accuracy": 0.8083000521564004,
+ "num_tokens": 1393922.0,
+ "step": 450
+ },
+ {
+ "entropy": 0.6749556151032448,
+ "epoch": 1.1683226183518411,
+ "grad_norm": 0.6444424390792847,
+ "learning_rate": 0.0002634191889632654,
+ "loss": 0.627038345336914,
+ "mean_token_accuracy": 0.8141268998384475,
+ "num_tokens": 1551442.0,
+ "step": 500
+ },
+ {
+ "entropy": 0.6727319830656051,
+ "epoch": 1.2852133255406195,
+ "grad_norm": 0.6360110640525818,
+ "learning_rate": 0.00026299877604786883,
+ "loss": 0.6232248687744141,
+ "mean_token_accuracy": 0.8153417313098907,
+ "num_tokens": 1706697.0,
+ "step": 550
+ },
+ {
+ "entropy": 0.6600116823613643,
+ "epoch": 1.4021040327293979,
+ "grad_norm": 0.46475180983543396,
+ "learning_rate": 0.00026236025574535523,
+ "loss": 0.6124750518798828,
+ "mean_token_accuracy": 0.8175293403863907,
+ "num_tokens": 1865246.0,
+ "step": 600
+ },
+ {
+ "entropy": 0.6494863349199295,
+ "epoch": 1.5189947399181765,
+ "grad_norm": 0.6132755875587463,
+ "learning_rate": 0.00026150468970660587,
+ "loss": 0.5997534942626953,
+ "mean_token_accuracy": 0.822050162255764,
+ "num_tokens": 2022060.0,
+ "step": 650
+ },
+ {
+ "entropy": 0.6511707258224487,
+ "epoch": 1.635885447106955,
+ "grad_norm": 0.570785641670227,
+ "learning_rate": 0.00026043350045872797,
+ "loss": 0.6059062576293945,
+ "mean_token_accuracy": 0.820236160159111,
+ "num_tokens": 2178909.0,
+ "step": 700
+ },
+ {
+ "entropy": 0.6550412529706955,
+ "epoch": 1.7527761542957334,
+ "grad_norm": 0.5398300290107727,
+ "learning_rate": 0.0002591484690398568,
+ "loss": 0.6031087875366211,
+ "mean_token_accuracy": 0.8206016793847084,
+ "num_tokens": 2329314.0,
+ "step": 750
+ },
+ {
+ "entropy": 0.6509851542115211,
+ "epoch": 1.869666861484512,
+ "grad_norm": 0.6913411617279053,
+ "learning_rate": 0.0002576517320378704,
+ "loss": 0.6023480224609375,
+ "mean_token_accuracy": 0.822216115295887,
+ "num_tokens": 2478559.0,
+ "step": 800
+ },
+ {
+ "entropy": 0.6560930109024048,
+ "epoch": 1.9865575686732906,
+ "grad_norm": 0.5568020939826965,
+ "learning_rate": 0.0002559457780379408,
+ "loss": 0.6031641006469727,
+ "mean_token_accuracy": 0.8205807811021805,
+ "num_tokens": 2631937.0,
+ "step": 850
+ },
+ {
+ "epoch": 2.0,
+ "eval_entropy": 0.794637631866839,
+ "eval_loss": 0.7390629649162292,
+ "eval_mean_token_accuracy": 0.7983863295672776,
+ "eval_num_tokens": 2651286.0,
+ "eval_runtime": 36.7346,
+ "eval_samples_per_second": 33.511,
+ "eval_steps_per_second": 4.192,
+ "step": 856
+ },
+ {
+ "entropy": 0.5709320336730037,
+ "epoch": 2.102863822326125,
+ "grad_norm": 0.4339260458946228,
+ "learning_rate": 0.0002540334434848298,
+ "loss": 0.5152394485473633,
+ "mean_token_accuracy": 0.8391561669919958,
+ "num_tokens": 2795200.0,
+ "step": 900
+ },
+ {
+ "entropy": 0.5768725140392781,
+ "epoch": 2.2197545295149035,
+ "grad_norm": 0.6345565319061279,
+ "learning_rate": 0.00025191790796680617,
+ "loss": 0.5241253662109375,
+ "mean_token_accuracy": 0.8377170553803444,
+ "num_tokens": 2950534.0,
+ "step": 950
+ },
+ {
+ "entropy": 0.570428695231676,
+ "epoch": 2.3366452367036823,
+ "grad_norm": 0.5838360786437988,
+ "learning_rate": 0.00024960268892902876,
+ "loss": 0.5187693405151367,
+ "mean_token_accuracy": 0.8382867723703384,
+ "num_tokens": 3101471.0,
+ "step": 1000
+ },
+ {
+ "entropy": 0.5713468404114246,
+ "epoch": 2.4535359438924607,
+ "grad_norm": 0.5641153454780579,
+ "learning_rate": 0.0002470916358251826,
+ "loss": 0.5234761810302735,
+ "mean_token_accuracy": 0.8382993021607399,
+ "num_tokens": 3258554.0,
+ "step": 1050
+ },
+ {
+ "entropy": 0.5645828676223755,
+ "epoch": 2.570426651081239,
+ "grad_norm": 0.504467785358429,
+ "learning_rate": 0.0002443889237170939,
+ "loss": 0.5156657409667968,
+ "mean_token_accuracy": 0.8391393664479255,
+ "num_tokens": 3416088.0,
+ "step": 1100
+ },
+ {
+ "entropy": 0.5659542319178581,
+ "epoch": 2.6873173582700174,
+ "grad_norm": 0.5225860476493835,
+ "learning_rate": 0.00024149904633296492,
+ "loss": 0.5180046844482422,
+ "mean_token_accuracy": 0.8389606699347496,
+ "num_tokens": 3567251.0,
+ "step": 1150
+ },
+ {
+ "entropy": 0.5713718670606613,
+ "epoch": 2.8042080654587958,
+ "grad_norm": 0.4986768066883087,
+ "learning_rate": 0.0002384268085957702,
+ "loss": 0.524637565612793,
+ "mean_token_accuracy": 0.8381215763092041,
+ "num_tokens": 3719464.0,
+ "step": 1200
+ },
+ {
+ "entropy": 0.5710382495820522,
+ "epoch": 2.9210987726475746,
+ "grad_norm": 0.5548401474952698,
+ "learning_rate": 0.00023517731863423788,
+ "loss": 0.5251939773559571,
+ "mean_token_accuracy": 0.8378310143947602,
+ "num_tokens": 3874423.0,
+ "step": 1250
+ },
+ {
+ "epoch": 3.0,
+ "eval_entropy": 0.6970683218209774,
+ "eval_loss": 0.7369498014450073,
+ "eval_mean_token_accuracy": 0.7919643076209278,
+ "eval_num_tokens": 3976929.0,
+ "eval_runtime": 36.9722,
+ "eval_samples_per_second": 33.295,
+ "eval_steps_per_second": 4.165,
+ "step": 1284
+ }
+ ],
+ "logging_steps": 50,
+ "max_steps": 4280,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 10,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 6.142907198829312e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1712/README.md b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1712/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1712/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1712/adapter_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1712/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9e45544de727016be134f25e606b7b2b87c67549
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1712/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.050044417519510255,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "v_proj",
+ "o_proj",
+ "q_proj",
+ "gate_proj",
+ "k_proj",
+ "down_proj",
+ "up_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1712/chat_template.jinja b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1712/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1712/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1712/tokenizer_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1712/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1712/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1712/trainer_state.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1712/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..bd1cdcabe854134d12b7611ddff01efda9f73c12
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-1712/trainer_state.json
@@ -0,0 +1,418 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 4.0,
+ "eval_steps": 500,
+ "global_step": 1712,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 1.8448095142841339,
+ "epoch": 0.11689070718877849,
+ "grad_norm": 1.4738068580627441,
+ "learning_rate": 3.0183098859435753e-05,
+ "loss": 1.6948321533203126,
+ "mean_token_accuracy": 0.6315939916670322,
+ "num_tokens": 153689.0,
+ "step": 50
+ },
+ {
+ "entropy": 0.9560942512750625,
+ "epoch": 0.23378141437755698,
+ "grad_norm": 1.378090739250183,
+ "learning_rate": 6.098217932824775e-05,
+ "loss": 0.8888477325439453,
+ "mean_token_accuracy": 0.7586692428588867,
+ "num_tokens": 310548.0,
+ "step": 100
+ },
+ {
+ "entropy": 0.848925698697567,
+ "epoch": 0.3506721215663355,
+ "grad_norm": 0.959621250629425,
+ "learning_rate": 9.178125979705974e-05,
+ "loss": 0.7850731658935547,
+ "mean_token_accuracy": 0.7774798944592476,
+ "num_tokens": 468560.0,
+ "step": 150
+ },
+ {
+ "entropy": 0.811761694252491,
+ "epoch": 0.46756282875511396,
+ "grad_norm": 0.8867514133453369,
+ "learning_rate": 0.00012258034026587175,
+ "loss": 0.7446884155273438,
+ "mean_token_accuracy": 0.7892421445250511,
+ "num_tokens": 619552.0,
+ "step": 200
+ },
+ {
+ "entropy": 0.7709847274422645,
+ "epoch": 0.5844535359438925,
+ "grad_norm": 0.5822142362594604,
+ "learning_rate": 0.00015337942073468373,
+ "loss": 0.7130480194091797,
+ "mean_token_accuracy": 0.7964023387432099,
+ "num_tokens": 771972.0,
+ "step": 250
+ },
+ {
+ "entropy": 0.7401632118225098,
+ "epoch": 0.701344243132671,
+ "grad_norm": 0.7557511925697327,
+ "learning_rate": 0.00018417850120349574,
+ "loss": 0.6857429504394531,
+ "mean_token_accuracy": 0.8041065946221352,
+ "num_tokens": 928088.0,
+ "step": 300
+ },
+ {
+ "entropy": 0.7295274111628532,
+ "epoch": 0.8182349503214494,
+ "grad_norm": 0.6769705414772034,
+ "learning_rate": 0.00021497758167230772,
+ "loss": 0.6801216125488281,
+ "mean_token_accuracy": 0.8037989619374275,
+ "num_tokens": 1086836.0,
+ "step": 350
+ },
+ {
+ "entropy": 0.710282272696495,
+ "epoch": 0.9351256575102279,
+ "grad_norm": 0.832009494304657,
+ "learning_rate": 0.00024577666214111967,
+ "loss": 0.6587177276611328,
+ "mean_token_accuracy": 0.8084024578332901,
+ "num_tokens": 1241227.0,
+ "step": 400
+ },
+ {
+ "epoch": 1.0,
+ "eval_entropy": 0.848237401672772,
+ "eval_loss": 0.7832363247871399,
+ "eval_mean_token_accuracy": 0.7817642971292719,
+ "eval_num_tokens": 1325643.0,
+ "eval_runtime": 36.8595,
+ "eval_samples_per_second": 33.397,
+ "eval_steps_per_second": 4.178,
+ "step": 428
+ },
+ {
+ "entropy": 0.7039205249230466,
+ "epoch": 1.0514319111630626,
+ "grad_norm": 0.6213153600692749,
+ "learning_rate": 0.00026362079548206905,
+ "loss": 0.6527813720703125,
+ "mean_token_accuracy": 0.8083000521564004,
+ "num_tokens": 1393922.0,
+ "step": 450
+ },
+ {
+ "entropy": 0.6749556151032448,
+ "epoch": 1.1683226183518411,
+ "grad_norm": 0.6444424390792847,
+ "learning_rate": 0.0002634191889632654,
+ "loss": 0.627038345336914,
+ "mean_token_accuracy": 0.8141268998384475,
+ "num_tokens": 1551442.0,
+ "step": 500
+ },
+ {
+ "entropy": 0.6727319830656051,
+ "epoch": 1.2852133255406195,
+ "grad_norm": 0.6360110640525818,
+ "learning_rate": 0.00026299877604786883,
+ "loss": 0.6232248687744141,
+ "mean_token_accuracy": 0.8153417313098907,
+ "num_tokens": 1706697.0,
+ "step": 550
+ },
+ {
+ "entropy": 0.6600116823613643,
+ "epoch": 1.4021040327293979,
+ "grad_norm": 0.46475180983543396,
+ "learning_rate": 0.00026236025574535523,
+ "loss": 0.6124750518798828,
+ "mean_token_accuracy": 0.8175293403863907,
+ "num_tokens": 1865246.0,
+ "step": 600
+ },
+ {
+ "entropy": 0.6494863349199295,
+ "epoch": 1.5189947399181765,
+ "grad_norm": 0.6132755875587463,
+ "learning_rate": 0.00026150468970660587,
+ "loss": 0.5997534942626953,
+ "mean_token_accuracy": 0.822050162255764,
+ "num_tokens": 2022060.0,
+ "step": 650
+ },
+ {
+ "entropy": 0.6511707258224487,
+ "epoch": 1.635885447106955,
+ "grad_norm": 0.570785641670227,
+ "learning_rate": 0.00026043350045872797,
+ "loss": 0.6059062576293945,
+ "mean_token_accuracy": 0.820236160159111,
+ "num_tokens": 2178909.0,
+ "step": 700
+ },
+ {
+ "entropy": 0.6550412529706955,
+ "epoch": 1.7527761542957334,
+ "grad_norm": 0.5398300290107727,
+ "learning_rate": 0.0002591484690398568,
+ "loss": 0.6031087875366211,
+ "mean_token_accuracy": 0.8206016793847084,
+ "num_tokens": 2329314.0,
+ "step": 750
+ },
+ {
+ "entropy": 0.6509851542115211,
+ "epoch": 1.869666861484512,
+ "grad_norm": 0.6913411617279053,
+ "learning_rate": 0.0002576517320378704,
+ "loss": 0.6023480224609375,
+ "mean_token_accuracy": 0.822216115295887,
+ "num_tokens": 2478559.0,
+ "step": 800
+ },
+ {
+ "entropy": 0.6560930109024048,
+ "epoch": 1.9865575686732906,
+ "grad_norm": 0.5568020939826965,
+ "learning_rate": 0.0002559457780379408,
+ "loss": 0.6031641006469727,
+ "mean_token_accuracy": 0.8205807811021805,
+ "num_tokens": 2631937.0,
+ "step": 850
+ },
+ {
+ "epoch": 2.0,
+ "eval_entropy": 0.794637631866839,
+ "eval_loss": 0.7390629649162292,
+ "eval_mean_token_accuracy": 0.7983863295672776,
+ "eval_num_tokens": 2651286.0,
+ "eval_runtime": 36.7346,
+ "eval_samples_per_second": 33.511,
+ "eval_steps_per_second": 4.192,
+ "step": 856
+ },
+ {
+ "entropy": 0.5709320336730037,
+ "epoch": 2.102863822326125,
+ "grad_norm": 0.4339260458946228,
+ "learning_rate": 0.0002540334434848298,
+ "loss": 0.5152394485473633,
+ "mean_token_accuracy": 0.8391561669919958,
+ "num_tokens": 2795200.0,
+ "step": 900
+ },
+ {
+ "entropy": 0.5768725140392781,
+ "epoch": 2.2197545295149035,
+ "grad_norm": 0.6345565319061279,
+ "learning_rate": 0.00025191790796680617,
+ "loss": 0.5241253662109375,
+ "mean_token_accuracy": 0.8377170553803444,
+ "num_tokens": 2950534.0,
+ "step": 950
+ },
+ {
+ "entropy": 0.570428695231676,
+ "epoch": 2.3366452367036823,
+ "grad_norm": 0.5838360786437988,
+ "learning_rate": 0.00024960268892902876,
+ "loss": 0.5187693405151367,
+ "mean_token_accuracy": 0.8382867723703384,
+ "num_tokens": 3101471.0,
+ "step": 1000
+ },
+ {
+ "entropy": 0.5713468404114246,
+ "epoch": 2.4535359438924607,
+ "grad_norm": 0.5641153454780579,
+ "learning_rate": 0.0002470916358251826,
+ "loss": 0.5234761810302735,
+ "mean_token_accuracy": 0.8382993021607399,
+ "num_tokens": 3258554.0,
+ "step": 1050
+ },
+ {
+ "entropy": 0.5645828676223755,
+ "epoch": 2.570426651081239,
+ "grad_norm": 0.504467785358429,
+ "learning_rate": 0.0002443889237170939,
+ "loss": 0.5156657409667968,
+ "mean_token_accuracy": 0.8391393664479255,
+ "num_tokens": 3416088.0,
+ "step": 1100
+ },
+ {
+ "entropy": 0.5659542319178581,
+ "epoch": 2.6873173582700174,
+ "grad_norm": 0.5225860476493835,
+ "learning_rate": 0.00024149904633296492,
+ "loss": 0.5180046844482422,
+ "mean_token_accuracy": 0.8389606699347496,
+ "num_tokens": 3567251.0,
+ "step": 1150
+ },
+ {
+ "entropy": 0.5713718670606613,
+ "epoch": 2.8042080654587958,
+ "grad_norm": 0.4986768066883087,
+ "learning_rate": 0.0002384268085957702,
+ "loss": 0.524637565612793,
+ "mean_token_accuracy": 0.8381215763092041,
+ "num_tokens": 3719464.0,
+ "step": 1200
+ },
+ {
+ "entropy": 0.5710382495820522,
+ "epoch": 2.9210987726475746,
+ "grad_norm": 0.5548401474952698,
+ "learning_rate": 0.00023517731863423788,
+ "loss": 0.5251939773559571,
+ "mean_token_accuracy": 0.8378310143947602,
+ "num_tokens": 3874423.0,
+ "step": 1250
+ },
+ {
+ "epoch": 3.0,
+ "eval_entropy": 0.6970683218209774,
+ "eval_loss": 0.7369498014450073,
+ "eval_mean_token_accuracy": 0.7919643076209278,
+ "eval_num_tokens": 3976929.0,
+ "eval_runtime": 36.9722,
+ "eval_samples_per_second": 33.295,
+ "eval_steps_per_second": 4.165,
+ "step": 1284
+ },
+ {
+ "entropy": 0.5407733828877684,
+ "epoch": 3.037405026300409,
+ "grad_norm": 0.6430732607841492,
+ "learning_rate": 0.00023175597928969878,
+ "loss": 0.49219757080078125,
+ "mean_token_accuracy": 0.8467060854686564,
+ "num_tokens": 4027609.0,
+ "step": 1300
+ },
+ {
+ "entropy": 0.4786584210395813,
+ "epoch": 3.1542957334891875,
+ "grad_norm": 0.6474543809890747,
+ "learning_rate": 0.00022816847913292434,
+ "loss": 0.4224985122680664,
+ "mean_token_accuracy": 0.8628773537278175,
+ "num_tokens": 4180560.0,
+ "step": 1350
+ },
+ {
+ "entropy": 0.49172899857163427,
+ "epoch": 3.2711864406779663,
+ "grad_norm": 0.549326479434967,
+ "learning_rate": 0.00022442078300588989,
+ "loss": 0.43718509674072265,
+ "mean_token_accuracy": 0.8583108323812485,
+ "num_tokens": 4329954.0,
+ "step": 1400
+ },
+ {
+ "entropy": 0.4829080794751644,
+ "epoch": 3.3880771478667446,
+ "grad_norm": 0.5274888873100281,
+ "learning_rate": 0.00022051912210418922,
+ "loss": 0.4315507125854492,
+ "mean_token_accuracy": 0.8612637943029404,
+ "num_tokens": 4484354.0,
+ "step": 1450
+ },
+ {
+ "entropy": 0.4855460473895073,
+ "epoch": 3.504967855055523,
+ "grad_norm": 0.5846306681632996,
+ "learning_rate": 0.00021646998361658985,
+ "loss": 0.43724521636962893,
+ "mean_token_accuracy": 0.8589270997047425,
+ "num_tokens": 4638919.0,
+ "step": 1500
+ },
+ {
+ "entropy": 0.4905838692188263,
+ "epoch": 3.6218585622443014,
+ "grad_norm": 0.5149123072624207,
+ "learning_rate": 0.00021228009993895495,
+ "loss": 0.44197372436523436,
+ "mean_token_accuracy": 0.8584243017435074,
+ "num_tokens": 4795540.0,
+ "step": 1550
+ },
+ {
+ "entropy": 0.4994237454235554,
+ "epoch": 3.73874926943308,
+ "grad_norm": 0.5417225956916809,
+ "learning_rate": 0.0002079564374804665,
+ "loss": 0.4491108322143555,
+ "mean_token_accuracy": 0.8556393891572952,
+ "num_tokens": 4951177.0,
+ "step": 1600
+ },
+ {
+ "entropy": 0.49903531357645986,
+ "epoch": 3.8556399766218585,
+ "grad_norm": 0.5086734890937805,
+ "learning_rate": 0.00020350618508076015,
+ "loss": 0.4518918228149414,
+ "mean_token_accuracy": 0.8549707370996475,
+ "num_tokens": 5104814.0,
+ "step": 1650
+ },
+ {
+ "entropy": 0.4859523306787014,
+ "epoch": 3.972530683810637,
+ "grad_norm": 0.5446717739105225,
+ "learning_rate": 0.00019893674205723094,
+ "loss": 0.4410313415527344,
+ "mean_token_accuracy": 0.8587654489278793,
+ "num_tokens": 5265185.0,
+ "step": 1700
+ },
+ {
+ "epoch": 4.0,
+ "eval_entropy": 0.6577693150415049,
+ "eval_loss": 0.7380754947662354,
+ "eval_mean_token_accuracy": 0.8039549565934515,
+ "eval_num_tokens": 5302572.0,
+ "eval_runtime": 36.5757,
+ "eval_samples_per_second": 33.656,
+ "eval_steps_per_second": 4.21,
+ "step": 1712
+ }
+ ],
+ "logging_steps": 50,
+ "max_steps": 4280,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 10,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 8.189719176112896e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2140/README.md b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2140/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2140/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2140/adapter_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2140/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9e45544de727016be134f25e606b7b2b87c67549
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2140/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.050044417519510255,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "v_proj",
+ "o_proj",
+ "q_proj",
+ "gate_proj",
+ "k_proj",
+ "down_proj",
+ "up_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2140/chat_template.jinja b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2140/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2140/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2140/tokenizer_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2140/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2140/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2140/trainer_state.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2140/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6928b20771048bd0277ae0671aca8b8ae240f925
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2140/trainer_state.json
@@ -0,0 +1,509 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 5.0,
+ "eval_steps": 500,
+ "global_step": 2140,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 1.8448095142841339,
+ "epoch": 0.11689070718877849,
+ "grad_norm": 1.4738068580627441,
+ "learning_rate": 3.0183098859435753e-05,
+ "loss": 1.6948321533203126,
+ "mean_token_accuracy": 0.6315939916670322,
+ "num_tokens": 153689.0,
+ "step": 50
+ },
+ {
+ "entropy": 0.9560942512750625,
+ "epoch": 0.23378141437755698,
+ "grad_norm": 1.378090739250183,
+ "learning_rate": 6.098217932824775e-05,
+ "loss": 0.8888477325439453,
+ "mean_token_accuracy": 0.7586692428588867,
+ "num_tokens": 310548.0,
+ "step": 100
+ },
+ {
+ "entropy": 0.848925698697567,
+ "epoch": 0.3506721215663355,
+ "grad_norm": 0.959621250629425,
+ "learning_rate": 9.178125979705974e-05,
+ "loss": 0.7850731658935547,
+ "mean_token_accuracy": 0.7774798944592476,
+ "num_tokens": 468560.0,
+ "step": 150
+ },
+ {
+ "entropy": 0.811761694252491,
+ "epoch": 0.46756282875511396,
+ "grad_norm": 0.8867514133453369,
+ "learning_rate": 0.00012258034026587175,
+ "loss": 0.7446884155273438,
+ "mean_token_accuracy": 0.7892421445250511,
+ "num_tokens": 619552.0,
+ "step": 200
+ },
+ {
+ "entropy": 0.7709847274422645,
+ "epoch": 0.5844535359438925,
+ "grad_norm": 0.5822142362594604,
+ "learning_rate": 0.00015337942073468373,
+ "loss": 0.7130480194091797,
+ "mean_token_accuracy": 0.7964023387432099,
+ "num_tokens": 771972.0,
+ "step": 250
+ },
+ {
+ "entropy": 0.7401632118225098,
+ "epoch": 0.701344243132671,
+ "grad_norm": 0.7557511925697327,
+ "learning_rate": 0.00018417850120349574,
+ "loss": 0.6857429504394531,
+ "mean_token_accuracy": 0.8041065946221352,
+ "num_tokens": 928088.0,
+ "step": 300
+ },
+ {
+ "entropy": 0.7295274111628532,
+ "epoch": 0.8182349503214494,
+ "grad_norm": 0.6769705414772034,
+ "learning_rate": 0.00021497758167230772,
+ "loss": 0.6801216125488281,
+ "mean_token_accuracy": 0.8037989619374275,
+ "num_tokens": 1086836.0,
+ "step": 350
+ },
+ {
+ "entropy": 0.710282272696495,
+ "epoch": 0.9351256575102279,
+ "grad_norm": 0.832009494304657,
+ "learning_rate": 0.00024577666214111967,
+ "loss": 0.6587177276611328,
+ "mean_token_accuracy": 0.8084024578332901,
+ "num_tokens": 1241227.0,
+ "step": 400
+ },
+ {
+ "epoch": 1.0,
+ "eval_entropy": 0.848237401672772,
+ "eval_loss": 0.7832363247871399,
+ "eval_mean_token_accuracy": 0.7817642971292719,
+ "eval_num_tokens": 1325643.0,
+ "eval_runtime": 36.8595,
+ "eval_samples_per_second": 33.397,
+ "eval_steps_per_second": 4.178,
+ "step": 428
+ },
+ {
+ "entropy": 0.7039205249230466,
+ "epoch": 1.0514319111630626,
+ "grad_norm": 0.6213153600692749,
+ "learning_rate": 0.00026362079548206905,
+ "loss": 0.6527813720703125,
+ "mean_token_accuracy": 0.8083000521564004,
+ "num_tokens": 1393922.0,
+ "step": 450
+ },
+ {
+ "entropy": 0.6749556151032448,
+ "epoch": 1.1683226183518411,
+ "grad_norm": 0.6444424390792847,
+ "learning_rate": 0.0002634191889632654,
+ "loss": 0.627038345336914,
+ "mean_token_accuracy": 0.8141268998384475,
+ "num_tokens": 1551442.0,
+ "step": 500
+ },
+ {
+ "entropy": 0.6727319830656051,
+ "epoch": 1.2852133255406195,
+ "grad_norm": 0.6360110640525818,
+ "learning_rate": 0.00026299877604786883,
+ "loss": 0.6232248687744141,
+ "mean_token_accuracy": 0.8153417313098907,
+ "num_tokens": 1706697.0,
+ "step": 550
+ },
+ {
+ "entropy": 0.6600116823613643,
+ "epoch": 1.4021040327293979,
+ "grad_norm": 0.46475180983543396,
+ "learning_rate": 0.00026236025574535523,
+ "loss": 0.6124750518798828,
+ "mean_token_accuracy": 0.8175293403863907,
+ "num_tokens": 1865246.0,
+ "step": 600
+ },
+ {
+ "entropy": 0.6494863349199295,
+ "epoch": 1.5189947399181765,
+ "grad_norm": 0.6132755875587463,
+ "learning_rate": 0.00026150468970660587,
+ "loss": 0.5997534942626953,
+ "mean_token_accuracy": 0.822050162255764,
+ "num_tokens": 2022060.0,
+ "step": 650
+ },
+ {
+ "entropy": 0.6511707258224487,
+ "epoch": 1.635885447106955,
+ "grad_norm": 0.570785641670227,
+ "learning_rate": 0.00026043350045872797,
+ "loss": 0.6059062576293945,
+ "mean_token_accuracy": 0.820236160159111,
+ "num_tokens": 2178909.0,
+ "step": 700
+ },
+ {
+ "entropy": 0.6550412529706955,
+ "epoch": 1.7527761542957334,
+ "grad_norm": 0.5398300290107727,
+ "learning_rate": 0.0002591484690398568,
+ "loss": 0.6031087875366211,
+ "mean_token_accuracy": 0.8206016793847084,
+ "num_tokens": 2329314.0,
+ "step": 750
+ },
+ {
+ "entropy": 0.6509851542115211,
+ "epoch": 1.869666861484512,
+ "grad_norm": 0.6913411617279053,
+ "learning_rate": 0.0002576517320378704,
+ "loss": 0.6023480224609375,
+ "mean_token_accuracy": 0.822216115295887,
+ "num_tokens": 2478559.0,
+ "step": 800
+ },
+ {
+ "entropy": 0.6560930109024048,
+ "epoch": 1.9865575686732906,
+ "grad_norm": 0.5568020939826965,
+ "learning_rate": 0.0002559457780379408,
+ "loss": 0.6031641006469727,
+ "mean_token_accuracy": 0.8205807811021805,
+ "num_tokens": 2631937.0,
+ "step": 850
+ },
+ {
+ "epoch": 2.0,
+ "eval_entropy": 0.794637631866839,
+ "eval_loss": 0.7390629649162292,
+ "eval_mean_token_accuracy": 0.7983863295672776,
+ "eval_num_tokens": 2651286.0,
+ "eval_runtime": 36.7346,
+ "eval_samples_per_second": 33.511,
+ "eval_steps_per_second": 4.192,
+ "step": 856
+ },
+ {
+ "entropy": 0.5709320336730037,
+ "epoch": 2.102863822326125,
+ "grad_norm": 0.4339260458946228,
+ "learning_rate": 0.0002540334434848298,
+ "loss": 0.5152394485473633,
+ "mean_token_accuracy": 0.8391561669919958,
+ "num_tokens": 2795200.0,
+ "step": 900
+ },
+ {
+ "entropy": 0.5768725140392781,
+ "epoch": 2.2197545295149035,
+ "grad_norm": 0.6345565319061279,
+ "learning_rate": 0.00025191790796680617,
+ "loss": 0.5241253662109375,
+ "mean_token_accuracy": 0.8377170553803444,
+ "num_tokens": 2950534.0,
+ "step": 950
+ },
+ {
+ "entropy": 0.570428695231676,
+ "epoch": 2.3366452367036823,
+ "grad_norm": 0.5838360786437988,
+ "learning_rate": 0.00024960268892902876,
+ "loss": 0.5187693405151367,
+ "mean_token_accuracy": 0.8382867723703384,
+ "num_tokens": 3101471.0,
+ "step": 1000
+ },
+ {
+ "entropy": 0.5713468404114246,
+ "epoch": 2.4535359438924607,
+ "grad_norm": 0.5641153454780579,
+ "learning_rate": 0.0002470916358251826,
+ "loss": 0.5234761810302735,
+ "mean_token_accuracy": 0.8382993021607399,
+ "num_tokens": 3258554.0,
+ "step": 1050
+ },
+ {
+ "entropy": 0.5645828676223755,
+ "epoch": 2.570426651081239,
+ "grad_norm": 0.504467785358429,
+ "learning_rate": 0.0002443889237170939,
+ "loss": 0.5156657409667968,
+ "mean_token_accuracy": 0.8391393664479255,
+ "num_tokens": 3416088.0,
+ "step": 1100
+ },
+ {
+ "entropy": 0.5659542319178581,
+ "epoch": 2.6873173582700174,
+ "grad_norm": 0.5225860476493835,
+ "learning_rate": 0.00024149904633296492,
+ "loss": 0.5180046844482422,
+ "mean_token_accuracy": 0.8389606699347496,
+ "num_tokens": 3567251.0,
+ "step": 1150
+ },
+ {
+ "entropy": 0.5713718670606613,
+ "epoch": 2.8042080654587958,
+ "grad_norm": 0.4986768066883087,
+ "learning_rate": 0.0002384268085957702,
+ "loss": 0.524637565612793,
+ "mean_token_accuracy": 0.8381215763092041,
+ "num_tokens": 3719464.0,
+ "step": 1200
+ },
+ {
+ "entropy": 0.5710382495820522,
+ "epoch": 2.9210987726475746,
+ "grad_norm": 0.5548401474952698,
+ "learning_rate": 0.00023517731863423788,
+ "loss": 0.5251939773559571,
+ "mean_token_accuracy": 0.8378310143947602,
+ "num_tokens": 3874423.0,
+ "step": 1250
+ },
+ {
+ "epoch": 3.0,
+ "eval_entropy": 0.6970683218209774,
+ "eval_loss": 0.7369498014450073,
+ "eval_mean_token_accuracy": 0.7919643076209278,
+ "eval_num_tokens": 3976929.0,
+ "eval_runtime": 36.9722,
+ "eval_samples_per_second": 33.295,
+ "eval_steps_per_second": 4.165,
+ "step": 1284
+ },
+ {
+ "entropy": 0.5407733828877684,
+ "epoch": 3.037405026300409,
+ "grad_norm": 0.6430732607841492,
+ "learning_rate": 0.00023175597928969878,
+ "loss": 0.49219757080078125,
+ "mean_token_accuracy": 0.8467060854686564,
+ "num_tokens": 4027609.0,
+ "step": 1300
+ },
+ {
+ "entropy": 0.4786584210395813,
+ "epoch": 3.1542957334891875,
+ "grad_norm": 0.6474543809890747,
+ "learning_rate": 0.00022816847913292434,
+ "loss": 0.4224985122680664,
+ "mean_token_accuracy": 0.8628773537278175,
+ "num_tokens": 4180560.0,
+ "step": 1350
+ },
+ {
+ "entropy": 0.49172899857163427,
+ "epoch": 3.2711864406779663,
+ "grad_norm": 0.549326479434967,
+ "learning_rate": 0.00022442078300588989,
+ "loss": 0.43718509674072265,
+ "mean_token_accuracy": 0.8583108323812485,
+ "num_tokens": 4329954.0,
+ "step": 1400
+ },
+ {
+ "entropy": 0.4829080794751644,
+ "epoch": 3.3880771478667446,
+ "grad_norm": 0.5274888873100281,
+ "learning_rate": 0.00022051912210418922,
+ "loss": 0.4315507125854492,
+ "mean_token_accuracy": 0.8612637943029404,
+ "num_tokens": 4484354.0,
+ "step": 1450
+ },
+ {
+ "entropy": 0.4855460473895073,
+ "epoch": 3.504967855055523,
+ "grad_norm": 0.5846306681632996,
+ "learning_rate": 0.00021646998361658985,
+ "loss": 0.43724521636962893,
+ "mean_token_accuracy": 0.8589270997047425,
+ "num_tokens": 4638919.0,
+ "step": 1500
+ },
+ {
+ "entropy": 0.4905838692188263,
+ "epoch": 3.6218585622443014,
+ "grad_norm": 0.5149123072624207,
+ "learning_rate": 0.00021228009993895495,
+ "loss": 0.44197372436523436,
+ "mean_token_accuracy": 0.8584243017435074,
+ "num_tokens": 4795540.0,
+ "step": 1550
+ },
+ {
+ "entropy": 0.4994237454235554,
+ "epoch": 3.73874926943308,
+ "grad_norm": 0.5417225956916809,
+ "learning_rate": 0.0002079564374804665,
+ "loss": 0.4491108322143555,
+ "mean_token_accuracy": 0.8556393891572952,
+ "num_tokens": 4951177.0,
+ "step": 1600
+ },
+ {
+ "entropy": 0.49903531357645986,
+ "epoch": 3.8556399766218585,
+ "grad_norm": 0.5086734890937805,
+ "learning_rate": 0.00020350618508076015,
+ "loss": 0.4518918228149414,
+ "mean_token_accuracy": 0.8549707370996475,
+ "num_tokens": 5104814.0,
+ "step": 1650
+ },
+ {
+ "entropy": 0.4859523306787014,
+ "epoch": 3.972530683810637,
+ "grad_norm": 0.5446717739105225,
+ "learning_rate": 0.00019893674205723094,
+ "loss": 0.4410313415527344,
+ "mean_token_accuracy": 0.8587654489278793,
+ "num_tokens": 5265185.0,
+ "step": 1700
+ },
+ {
+ "epoch": 4.0,
+ "eval_entropy": 0.6577693150415049,
+ "eval_loss": 0.7380754947662354,
+ "eval_mean_token_accuracy": 0.8039549565934515,
+ "eval_num_tokens": 5302572.0,
+ "eval_runtime": 36.5757,
+ "eval_samples_per_second": 33.656,
+ "eval_steps_per_second": 4.21,
+ "step": 1712
+ },
+ {
+ "entropy": 0.41945636961328325,
+ "epoch": 4.088836937463472,
+ "grad_norm": 0.5925881266593933,
+ "learning_rate": 0.00019425570590238332,
+ "loss": 0.3575031280517578,
+ "mean_token_accuracy": 0.8812823918596584,
+ "num_tokens": 5419002.0,
+ "step": 1750
+ },
+ {
+ "entropy": 0.40106018364429474,
+ "epoch": 4.20572764465225,
+ "grad_norm": 0.7094829082489014,
+ "learning_rate": 0.0001894708596516806,
+ "loss": 0.3372145080566406,
+ "mean_token_accuracy": 0.8856011089682579,
+ "num_tokens": 5572969.0,
+ "step": 1800
+ },
+ {
+ "entropy": 0.4121378688514233,
+ "epoch": 4.322618351841029,
+ "grad_norm": 0.6077721118927002,
+ "learning_rate": 0.00018459015894289695,
+ "loss": 0.34956165313720705,
+ "mean_token_accuracy": 0.8818695649504662,
+ "num_tokens": 5733805.0,
+ "step": 1850
+ },
+ {
+ "entropy": 0.40683484479784965,
+ "epoch": 4.439509059029807,
+ "grad_norm": 0.5614883303642273,
+ "learning_rate": 0.00017962171878848807,
+ "loss": 0.34749691009521483,
+ "mean_token_accuracy": 0.882930768430233,
+ "num_tokens": 5888694.0,
+ "step": 1900
+ },
+ {
+ "entropy": 0.4100585526227951,
+ "epoch": 4.556399766218585,
+ "grad_norm": 0.5506443977355957,
+ "learning_rate": 0.00017457380008297383,
+ "loss": 0.3501173782348633,
+ "mean_token_accuracy": 0.8827524068951607,
+ "num_tokens": 6042655.0,
+ "step": 1950
+ },
+ {
+ "entropy": 0.41173958137631417,
+ "epoch": 4.673290473407365,
+ "grad_norm": 0.5409566164016724,
+ "learning_rate": 0.00016945479586776672,
+ "loss": 0.3541391754150391,
+ "mean_token_accuracy": 0.8808356913924217,
+ "num_tokens": 6203980.0,
+ "step": 2000
+ },
+ {
+ "entropy": 0.4158783262968063,
+ "epoch": 4.790181180596143,
+ "grad_norm": 0.572618842124939,
+ "learning_rate": 0.00016427321737628295,
+ "loss": 0.3536404800415039,
+ "mean_token_accuracy": 0.880456589460373,
+ "num_tokens": 6354604.0,
+ "step": 2050
+ },
+ {
+ "entropy": 0.408166281580925,
+ "epoch": 4.907071887784921,
+ "grad_norm": 0.5987762808799744,
+ "learning_rate": 0.00015903767988253897,
+ "loss": 0.3518599319458008,
+ "mean_token_accuracy": 0.8815453574061394,
+ "num_tokens": 6507484.0,
+ "step": 2100
+ },
+ {
+ "epoch": 5.0,
+ "eval_entropy": 0.6047254425364655,
+ "eval_loss": 0.7816920876502991,
+ "eval_mean_token_accuracy": 0.7987685733801358,
+ "eval_num_tokens": 6628215.0,
+ "eval_runtime": 36.9374,
+ "eval_samples_per_second": 33.327,
+ "eval_steps_per_second": 4.169,
+ "step": 2140
+ }
+ ],
+ "logging_steps": 50,
+ "max_steps": 4280,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 10,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.0236345370205184e+17,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2568/README.md b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2568/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2568/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2568/adapter_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2568/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9e45544de727016be134f25e606b7b2b87c67549
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2568/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.050044417519510255,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "v_proj",
+ "o_proj",
+ "q_proj",
+ "gate_proj",
+ "k_proj",
+ "down_proj",
+ "up_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2568/chat_template.jinja b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2568/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2568/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2568/tokenizer_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2568/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2568/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2568/trainer_state.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2568/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..50bb6eedd8378dfc2a02b04a5e958e3882db7456
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2568/trainer_state.json
@@ -0,0 +1,610 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 6.0,
+ "eval_steps": 500,
+ "global_step": 2568,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 1.8448095142841339,
+ "epoch": 0.11689070718877849,
+ "grad_norm": 1.4738068580627441,
+ "learning_rate": 3.0183098859435753e-05,
+ "loss": 1.6948321533203126,
+ "mean_token_accuracy": 0.6315939916670322,
+ "num_tokens": 153689.0,
+ "step": 50
+ },
+ {
+ "entropy": 0.9560942512750625,
+ "epoch": 0.23378141437755698,
+ "grad_norm": 1.378090739250183,
+ "learning_rate": 6.098217932824775e-05,
+ "loss": 0.8888477325439453,
+ "mean_token_accuracy": 0.7586692428588867,
+ "num_tokens": 310548.0,
+ "step": 100
+ },
+ {
+ "entropy": 0.848925698697567,
+ "epoch": 0.3506721215663355,
+ "grad_norm": 0.959621250629425,
+ "learning_rate": 9.178125979705974e-05,
+ "loss": 0.7850731658935547,
+ "mean_token_accuracy": 0.7774798944592476,
+ "num_tokens": 468560.0,
+ "step": 150
+ },
+ {
+ "entropy": 0.811761694252491,
+ "epoch": 0.46756282875511396,
+ "grad_norm": 0.8867514133453369,
+ "learning_rate": 0.00012258034026587175,
+ "loss": 0.7446884155273438,
+ "mean_token_accuracy": 0.7892421445250511,
+ "num_tokens": 619552.0,
+ "step": 200
+ },
+ {
+ "entropy": 0.7709847274422645,
+ "epoch": 0.5844535359438925,
+ "grad_norm": 0.5822142362594604,
+ "learning_rate": 0.00015337942073468373,
+ "loss": 0.7130480194091797,
+ "mean_token_accuracy": 0.7964023387432099,
+ "num_tokens": 771972.0,
+ "step": 250
+ },
+ {
+ "entropy": 0.7401632118225098,
+ "epoch": 0.701344243132671,
+ "grad_norm": 0.7557511925697327,
+ "learning_rate": 0.00018417850120349574,
+ "loss": 0.6857429504394531,
+ "mean_token_accuracy": 0.8041065946221352,
+ "num_tokens": 928088.0,
+ "step": 300
+ },
+ {
+ "entropy": 0.7295274111628532,
+ "epoch": 0.8182349503214494,
+ "grad_norm": 0.6769705414772034,
+ "learning_rate": 0.00021497758167230772,
+ "loss": 0.6801216125488281,
+ "mean_token_accuracy": 0.8037989619374275,
+ "num_tokens": 1086836.0,
+ "step": 350
+ },
+ {
+ "entropy": 0.710282272696495,
+ "epoch": 0.9351256575102279,
+ "grad_norm": 0.832009494304657,
+ "learning_rate": 0.00024577666214111967,
+ "loss": 0.6587177276611328,
+ "mean_token_accuracy": 0.8084024578332901,
+ "num_tokens": 1241227.0,
+ "step": 400
+ },
+ {
+ "epoch": 1.0,
+ "eval_entropy": 0.848237401672772,
+ "eval_loss": 0.7832363247871399,
+ "eval_mean_token_accuracy": 0.7817642971292719,
+ "eval_num_tokens": 1325643.0,
+ "eval_runtime": 36.8595,
+ "eval_samples_per_second": 33.397,
+ "eval_steps_per_second": 4.178,
+ "step": 428
+ },
+ {
+ "entropy": 0.7039205249230466,
+ "epoch": 1.0514319111630626,
+ "grad_norm": 0.6213153600692749,
+ "learning_rate": 0.00026362079548206905,
+ "loss": 0.6527813720703125,
+ "mean_token_accuracy": 0.8083000521564004,
+ "num_tokens": 1393922.0,
+ "step": 450
+ },
+ {
+ "entropy": 0.6749556151032448,
+ "epoch": 1.1683226183518411,
+ "grad_norm": 0.6444424390792847,
+ "learning_rate": 0.0002634191889632654,
+ "loss": 0.627038345336914,
+ "mean_token_accuracy": 0.8141268998384475,
+ "num_tokens": 1551442.0,
+ "step": 500
+ },
+ {
+ "entropy": 0.6727319830656051,
+ "epoch": 1.2852133255406195,
+ "grad_norm": 0.6360110640525818,
+ "learning_rate": 0.00026299877604786883,
+ "loss": 0.6232248687744141,
+ "mean_token_accuracy": 0.8153417313098907,
+ "num_tokens": 1706697.0,
+ "step": 550
+ },
+ {
+ "entropy": 0.6600116823613643,
+ "epoch": 1.4021040327293979,
+ "grad_norm": 0.46475180983543396,
+ "learning_rate": 0.00026236025574535523,
+ "loss": 0.6124750518798828,
+ "mean_token_accuracy": 0.8175293403863907,
+ "num_tokens": 1865246.0,
+ "step": 600
+ },
+ {
+ "entropy": 0.6494863349199295,
+ "epoch": 1.5189947399181765,
+ "grad_norm": 0.6132755875587463,
+ "learning_rate": 0.00026150468970660587,
+ "loss": 0.5997534942626953,
+ "mean_token_accuracy": 0.822050162255764,
+ "num_tokens": 2022060.0,
+ "step": 650
+ },
+ {
+ "entropy": 0.6511707258224487,
+ "epoch": 1.635885447106955,
+ "grad_norm": 0.570785641670227,
+ "learning_rate": 0.00026043350045872797,
+ "loss": 0.6059062576293945,
+ "mean_token_accuracy": 0.820236160159111,
+ "num_tokens": 2178909.0,
+ "step": 700
+ },
+ {
+ "entropy": 0.6550412529706955,
+ "epoch": 1.7527761542957334,
+ "grad_norm": 0.5398300290107727,
+ "learning_rate": 0.0002591484690398568,
+ "loss": 0.6031087875366211,
+ "mean_token_accuracy": 0.8206016793847084,
+ "num_tokens": 2329314.0,
+ "step": 750
+ },
+ {
+ "entropy": 0.6509851542115211,
+ "epoch": 1.869666861484512,
+ "grad_norm": 0.6913411617279053,
+ "learning_rate": 0.0002576517320378704,
+ "loss": 0.6023480224609375,
+ "mean_token_accuracy": 0.822216115295887,
+ "num_tokens": 2478559.0,
+ "step": 800
+ },
+ {
+ "entropy": 0.6560930109024048,
+ "epoch": 1.9865575686732906,
+ "grad_norm": 0.5568020939826965,
+ "learning_rate": 0.0002559457780379408,
+ "loss": 0.6031641006469727,
+ "mean_token_accuracy": 0.8205807811021805,
+ "num_tokens": 2631937.0,
+ "step": 850
+ },
+ {
+ "epoch": 2.0,
+ "eval_entropy": 0.794637631866839,
+ "eval_loss": 0.7390629649162292,
+ "eval_mean_token_accuracy": 0.7983863295672776,
+ "eval_num_tokens": 2651286.0,
+ "eval_runtime": 36.7346,
+ "eval_samples_per_second": 33.511,
+ "eval_steps_per_second": 4.192,
+ "step": 856
+ },
+ {
+ "entropy": 0.5709320336730037,
+ "epoch": 2.102863822326125,
+ "grad_norm": 0.4339260458946228,
+ "learning_rate": 0.0002540334434848298,
+ "loss": 0.5152394485473633,
+ "mean_token_accuracy": 0.8391561669919958,
+ "num_tokens": 2795200.0,
+ "step": 900
+ },
+ {
+ "entropy": 0.5768725140392781,
+ "epoch": 2.2197545295149035,
+ "grad_norm": 0.6345565319061279,
+ "learning_rate": 0.00025191790796680617,
+ "loss": 0.5241253662109375,
+ "mean_token_accuracy": 0.8377170553803444,
+ "num_tokens": 2950534.0,
+ "step": 950
+ },
+ {
+ "entropy": 0.570428695231676,
+ "epoch": 2.3366452367036823,
+ "grad_norm": 0.5838360786437988,
+ "learning_rate": 0.00024960268892902876,
+ "loss": 0.5187693405151367,
+ "mean_token_accuracy": 0.8382867723703384,
+ "num_tokens": 3101471.0,
+ "step": 1000
+ },
+ {
+ "entropy": 0.5713468404114246,
+ "epoch": 2.4535359438924607,
+ "grad_norm": 0.5641153454780579,
+ "learning_rate": 0.0002470916358251826,
+ "loss": 0.5234761810302735,
+ "mean_token_accuracy": 0.8382993021607399,
+ "num_tokens": 3258554.0,
+ "step": 1050
+ },
+ {
+ "entropy": 0.5645828676223755,
+ "epoch": 2.570426651081239,
+ "grad_norm": 0.504467785358429,
+ "learning_rate": 0.0002443889237170939,
+ "loss": 0.5156657409667968,
+ "mean_token_accuracy": 0.8391393664479255,
+ "num_tokens": 3416088.0,
+ "step": 1100
+ },
+ {
+ "entropy": 0.5659542319178581,
+ "epoch": 2.6873173582700174,
+ "grad_norm": 0.5225860476493835,
+ "learning_rate": 0.00024149904633296492,
+ "loss": 0.5180046844482422,
+ "mean_token_accuracy": 0.8389606699347496,
+ "num_tokens": 3567251.0,
+ "step": 1150
+ },
+ {
+ "entropy": 0.5713718670606613,
+ "epoch": 2.8042080654587958,
+ "grad_norm": 0.4986768066883087,
+ "learning_rate": 0.0002384268085957702,
+ "loss": 0.524637565612793,
+ "mean_token_accuracy": 0.8381215763092041,
+ "num_tokens": 3719464.0,
+ "step": 1200
+ },
+ {
+ "entropy": 0.5710382495820522,
+ "epoch": 2.9210987726475746,
+ "grad_norm": 0.5548401474952698,
+ "learning_rate": 0.00023517731863423788,
+ "loss": 0.5251939773559571,
+ "mean_token_accuracy": 0.8378310143947602,
+ "num_tokens": 3874423.0,
+ "step": 1250
+ },
+ {
+ "epoch": 3.0,
+ "eval_entropy": 0.6970683218209774,
+ "eval_loss": 0.7369498014450073,
+ "eval_mean_token_accuracy": 0.7919643076209278,
+ "eval_num_tokens": 3976929.0,
+ "eval_runtime": 36.9722,
+ "eval_samples_per_second": 33.295,
+ "eval_steps_per_second": 4.165,
+ "step": 1284
+ },
+ {
+ "entropy": 0.5407733828877684,
+ "epoch": 3.037405026300409,
+ "grad_norm": 0.6430732607841492,
+ "learning_rate": 0.00023175597928969878,
+ "loss": 0.49219757080078125,
+ "mean_token_accuracy": 0.8467060854686564,
+ "num_tokens": 4027609.0,
+ "step": 1300
+ },
+ {
+ "entropy": 0.4786584210395813,
+ "epoch": 3.1542957334891875,
+ "grad_norm": 0.6474543809890747,
+ "learning_rate": 0.00022816847913292434,
+ "loss": 0.4224985122680664,
+ "mean_token_accuracy": 0.8628773537278175,
+ "num_tokens": 4180560.0,
+ "step": 1350
+ },
+ {
+ "entropy": 0.49172899857163427,
+ "epoch": 3.2711864406779663,
+ "grad_norm": 0.549326479434967,
+ "learning_rate": 0.00022442078300588989,
+ "loss": 0.43718509674072265,
+ "mean_token_accuracy": 0.8583108323812485,
+ "num_tokens": 4329954.0,
+ "step": 1400
+ },
+ {
+ "entropy": 0.4829080794751644,
+ "epoch": 3.3880771478667446,
+ "grad_norm": 0.5274888873100281,
+ "learning_rate": 0.00022051912210418922,
+ "loss": 0.4315507125854492,
+ "mean_token_accuracy": 0.8612637943029404,
+ "num_tokens": 4484354.0,
+ "step": 1450
+ },
+ {
+ "entropy": 0.4855460473895073,
+ "epoch": 3.504967855055523,
+ "grad_norm": 0.5846306681632996,
+ "learning_rate": 0.00021646998361658985,
+ "loss": 0.43724521636962893,
+ "mean_token_accuracy": 0.8589270997047425,
+ "num_tokens": 4638919.0,
+ "step": 1500
+ },
+ {
+ "entropy": 0.4905838692188263,
+ "epoch": 3.6218585622443014,
+ "grad_norm": 0.5149123072624207,
+ "learning_rate": 0.00021228009993895495,
+ "loss": 0.44197372436523436,
+ "mean_token_accuracy": 0.8584243017435074,
+ "num_tokens": 4795540.0,
+ "step": 1550
+ },
+ {
+ "entropy": 0.4994237454235554,
+ "epoch": 3.73874926943308,
+ "grad_norm": 0.5417225956916809,
+ "learning_rate": 0.0002079564374804665,
+ "loss": 0.4491108322143555,
+ "mean_token_accuracy": 0.8556393891572952,
+ "num_tokens": 4951177.0,
+ "step": 1600
+ },
+ {
+ "entropy": 0.49903531357645986,
+ "epoch": 3.8556399766218585,
+ "grad_norm": 0.5086734890937805,
+ "learning_rate": 0.00020350618508076015,
+ "loss": 0.4518918228149414,
+ "mean_token_accuracy": 0.8549707370996475,
+ "num_tokens": 5104814.0,
+ "step": 1650
+ },
+ {
+ "entropy": 0.4859523306787014,
+ "epoch": 3.972530683810637,
+ "grad_norm": 0.5446717739105225,
+ "learning_rate": 0.00019893674205723094,
+ "loss": 0.4410313415527344,
+ "mean_token_accuracy": 0.8587654489278793,
+ "num_tokens": 5265185.0,
+ "step": 1700
+ },
+ {
+ "epoch": 4.0,
+ "eval_entropy": 0.6577693150415049,
+ "eval_loss": 0.7380754947662354,
+ "eval_mean_token_accuracy": 0.8039549565934515,
+ "eval_num_tokens": 5302572.0,
+ "eval_runtime": 36.5757,
+ "eval_samples_per_second": 33.656,
+ "eval_steps_per_second": 4.21,
+ "step": 1712
+ },
+ {
+ "entropy": 0.41945636961328325,
+ "epoch": 4.088836937463472,
+ "grad_norm": 0.5925881266593933,
+ "learning_rate": 0.00019425570590238332,
+ "loss": 0.3575031280517578,
+ "mean_token_accuracy": 0.8812823918596584,
+ "num_tokens": 5419002.0,
+ "step": 1750
+ },
+ {
+ "entropy": 0.40106018364429474,
+ "epoch": 4.20572764465225,
+ "grad_norm": 0.7094829082489014,
+ "learning_rate": 0.0001894708596516806,
+ "loss": 0.3372145080566406,
+ "mean_token_accuracy": 0.8856011089682579,
+ "num_tokens": 5572969.0,
+ "step": 1800
+ },
+ {
+ "entropy": 0.4121378688514233,
+ "epoch": 4.322618351841029,
+ "grad_norm": 0.6077721118927002,
+ "learning_rate": 0.00018459015894289695,
+ "loss": 0.34956165313720705,
+ "mean_token_accuracy": 0.8818695649504662,
+ "num_tokens": 5733805.0,
+ "step": 1850
+ },
+ {
+ "entropy": 0.40683484479784965,
+ "epoch": 4.439509059029807,
+ "grad_norm": 0.5614883303642273,
+ "learning_rate": 0.00017962171878848807,
+ "loss": 0.34749691009521483,
+ "mean_token_accuracy": 0.882930768430233,
+ "num_tokens": 5888694.0,
+ "step": 1900
+ },
+ {
+ "entropy": 0.4100585526227951,
+ "epoch": 4.556399766218585,
+ "grad_norm": 0.5506443977355957,
+ "learning_rate": 0.00017457380008297383,
+ "loss": 0.3501173782348633,
+ "mean_token_accuracy": 0.8827524068951607,
+ "num_tokens": 6042655.0,
+ "step": 1950
+ },
+ {
+ "entropy": 0.41173958137631417,
+ "epoch": 4.673290473407365,
+ "grad_norm": 0.5409566164016724,
+ "learning_rate": 0.00016945479586776672,
+ "loss": 0.3541391754150391,
+ "mean_token_accuracy": 0.8808356913924217,
+ "num_tokens": 6203980.0,
+ "step": 2000
+ },
+ {
+ "entropy": 0.4158783262968063,
+ "epoch": 4.790181180596143,
+ "grad_norm": 0.572618842124939,
+ "learning_rate": 0.00016427321737628295,
+ "loss": 0.3536404800415039,
+ "mean_token_accuracy": 0.880456589460373,
+ "num_tokens": 6354604.0,
+ "step": 2050
+ },
+ {
+ "entropy": 0.408166281580925,
+ "epoch": 4.907071887784921,
+ "grad_norm": 0.5987762808799744,
+ "learning_rate": 0.00015903767988253897,
+ "loss": 0.3518599319458008,
+ "mean_token_accuracy": 0.8815453574061394,
+ "num_tokens": 6507484.0,
+ "step": 2100
+ },
+ {
+ "epoch": 5.0,
+ "eval_entropy": 0.6047254425364655,
+ "eval_loss": 0.7816920876502991,
+ "eval_mean_token_accuracy": 0.7987685733801358,
+ "eval_num_tokens": 6628215.0,
+ "eval_runtime": 36.9374,
+ "eval_samples_per_second": 33.327,
+ "eval_steps_per_second": 4.169,
+ "step": 2140
+ },
+ {
+ "entropy": 0.38985791001188097,
+ "epoch": 5.023378141437756,
+ "grad_norm": 0.7454882860183716,
+ "learning_rate": 0.00015375688837676246,
+ "loss": 0.32916854858398437,
+ "mean_token_accuracy": 0.8886270373310875,
+ "num_tokens": 6662015.0,
+ "step": 2150
+ },
+ {
+ "entropy": 0.30170531339943407,
+ "epoch": 5.140268848626534,
+ "grad_norm": 0.5329811573028564,
+ "learning_rate": 0.0001484396230918344,
+ "loss": 0.23112209320068358,
+ "mean_token_accuracy": 0.9198204749822616,
+ "num_tokens": 6817989.0,
+ "step": 2200
+ },
+ {
+ "entropy": 0.3075008810311556,
+ "epoch": 5.257159555815313,
+ "grad_norm": 0.6896093487739563,
+ "learning_rate": 0.00014309472490462715,
+ "loss": 0.2416652488708496,
+ "mean_token_accuracy": 0.9158366405963898,
+ "num_tokens": 6977021.0,
+ "step": 2250
+ },
+ {
+ "entropy": 0.31153449676930905,
+ "epoch": 5.374050263004091,
+ "grad_norm": 0.6930833458900452,
+ "learning_rate": 0.00013773108063651161,
+ "loss": 0.24753889083862304,
+ "mean_token_accuracy": 0.9136614826321602,
+ "num_tokens": 7133360.0,
+ "step": 2300
+ },
+ {
+ "entropy": 0.3154346436262131,
+ "epoch": 5.490940970192869,
+ "grad_norm": 0.6476582884788513,
+ "learning_rate": 0.0001323576082774734,
+ "loss": 0.24872392654418946,
+ "mean_token_accuracy": 0.9135914054512978,
+ "num_tokens": 7282372.0,
+ "step": 2350
+ },
+ {
+ "entropy": 0.313641587048769,
+ "epoch": 5.607831677381649,
+ "grad_norm": 0.6953482627868652,
+ "learning_rate": 0.0001269832421584059,
+ "loss": 0.24919464111328124,
+ "mean_token_accuracy": 0.9131662029027939,
+ "num_tokens": 7435766.0,
+ "step": 2400
+ },
+ {
+ "entropy": 0.3163894522190094,
+ "epoch": 5.724722384570427,
+ "grad_norm": 0.6673940420150757,
+ "learning_rate": 0.00012161691809623384,
+ "loss": 0.25111726760864256,
+ "mean_token_accuracy": 0.913602537214756,
+ "num_tokens": 7578408.0,
+ "step": 2450
+ },
+ {
+ "entropy": 0.3183368914574385,
+ "epoch": 5.841613091759205,
+ "grad_norm": 0.7174487709999084,
+ "learning_rate": 0.00011626755853656576,
+ "loss": 0.2518378448486328,
+ "mean_token_accuracy": 0.912000589966774,
+ "num_tokens": 7734599.0,
+ "step": 2500
+ },
+ {
+ "entropy": 0.3077564483135939,
+ "epoch": 5.958503798947984,
+ "grad_norm": 0.6662200689315796,
+ "learning_rate": 0.00011094405771857891,
+ "loss": 0.24354087829589843,
+ "mean_token_accuracy": 0.9140791621804237,
+ "num_tokens": 7898028.0,
+ "step": 2550
+ },
+ {
+ "epoch": 6.0,
+ "eval_entropy": 0.5403795917699863,
+ "eval_loss": 0.8572548627853394,
+ "eval_mean_token_accuracy": 0.796143418395674,
+ "eval_num_tokens": 7953858.0,
+ "eval_runtime": 36.7176,
+ "eval_samples_per_second": 33.526,
+ "eval_steps_per_second": 4.194,
+ "step": 2568
+ }
+ ],
+ "logging_steps": 50,
+ "max_steps": 4280,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 10,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.228067205913728e+17,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2996/README.md b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2996/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2996/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2996/adapter_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2996/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9e45544de727016be134f25e606b7b2b87c67549
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2996/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.050044417519510255,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "v_proj",
+ "o_proj",
+ "q_proj",
+ "gate_proj",
+ "k_proj",
+ "down_proj",
+ "up_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2996/chat_template.jinja b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2996/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2996/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2996/tokenizer_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2996/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2996/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2996/trainer_state.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2996/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..0538ba571eaece25d59271908a07300dca6ecb5c
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-2996/trainer_state.json
@@ -0,0 +1,701 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 7.0,
+ "eval_steps": 500,
+ "global_step": 2996,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 1.8448095142841339,
+ "epoch": 0.11689070718877849,
+ "grad_norm": 1.4738068580627441,
+ "learning_rate": 3.0183098859435753e-05,
+ "loss": 1.6948321533203126,
+ "mean_token_accuracy": 0.6315939916670322,
+ "num_tokens": 153689.0,
+ "step": 50
+ },
+ {
+ "entropy": 0.9560942512750625,
+ "epoch": 0.23378141437755698,
+ "grad_norm": 1.378090739250183,
+ "learning_rate": 6.098217932824775e-05,
+ "loss": 0.8888477325439453,
+ "mean_token_accuracy": 0.7586692428588867,
+ "num_tokens": 310548.0,
+ "step": 100
+ },
+ {
+ "entropy": 0.848925698697567,
+ "epoch": 0.3506721215663355,
+ "grad_norm": 0.959621250629425,
+ "learning_rate": 9.178125979705974e-05,
+ "loss": 0.7850731658935547,
+ "mean_token_accuracy": 0.7774798944592476,
+ "num_tokens": 468560.0,
+ "step": 150
+ },
+ {
+ "entropy": 0.811761694252491,
+ "epoch": 0.46756282875511396,
+ "grad_norm": 0.8867514133453369,
+ "learning_rate": 0.00012258034026587175,
+ "loss": 0.7446884155273438,
+ "mean_token_accuracy": 0.7892421445250511,
+ "num_tokens": 619552.0,
+ "step": 200
+ },
+ {
+ "entropy": 0.7709847274422645,
+ "epoch": 0.5844535359438925,
+ "grad_norm": 0.5822142362594604,
+ "learning_rate": 0.00015337942073468373,
+ "loss": 0.7130480194091797,
+ "mean_token_accuracy": 0.7964023387432099,
+ "num_tokens": 771972.0,
+ "step": 250
+ },
+ {
+ "entropy": 0.7401632118225098,
+ "epoch": 0.701344243132671,
+ "grad_norm": 0.7557511925697327,
+ "learning_rate": 0.00018417850120349574,
+ "loss": 0.6857429504394531,
+ "mean_token_accuracy": 0.8041065946221352,
+ "num_tokens": 928088.0,
+ "step": 300
+ },
+ {
+ "entropy": 0.7295274111628532,
+ "epoch": 0.8182349503214494,
+ "grad_norm": 0.6769705414772034,
+ "learning_rate": 0.00021497758167230772,
+ "loss": 0.6801216125488281,
+ "mean_token_accuracy": 0.8037989619374275,
+ "num_tokens": 1086836.0,
+ "step": 350
+ },
+ {
+ "entropy": 0.710282272696495,
+ "epoch": 0.9351256575102279,
+ "grad_norm": 0.832009494304657,
+ "learning_rate": 0.00024577666214111967,
+ "loss": 0.6587177276611328,
+ "mean_token_accuracy": 0.8084024578332901,
+ "num_tokens": 1241227.0,
+ "step": 400
+ },
+ {
+ "epoch": 1.0,
+ "eval_entropy": 0.848237401672772,
+ "eval_loss": 0.7832363247871399,
+ "eval_mean_token_accuracy": 0.7817642971292719,
+ "eval_num_tokens": 1325643.0,
+ "eval_runtime": 36.8595,
+ "eval_samples_per_second": 33.397,
+ "eval_steps_per_second": 4.178,
+ "step": 428
+ },
+ {
+ "entropy": 0.7039205249230466,
+ "epoch": 1.0514319111630626,
+ "grad_norm": 0.6213153600692749,
+ "learning_rate": 0.00026362079548206905,
+ "loss": 0.6527813720703125,
+ "mean_token_accuracy": 0.8083000521564004,
+ "num_tokens": 1393922.0,
+ "step": 450
+ },
+ {
+ "entropy": 0.6749556151032448,
+ "epoch": 1.1683226183518411,
+ "grad_norm": 0.6444424390792847,
+ "learning_rate": 0.0002634191889632654,
+ "loss": 0.627038345336914,
+ "mean_token_accuracy": 0.8141268998384475,
+ "num_tokens": 1551442.0,
+ "step": 500
+ },
+ {
+ "entropy": 0.6727319830656051,
+ "epoch": 1.2852133255406195,
+ "grad_norm": 0.6360110640525818,
+ "learning_rate": 0.00026299877604786883,
+ "loss": 0.6232248687744141,
+ "mean_token_accuracy": 0.8153417313098907,
+ "num_tokens": 1706697.0,
+ "step": 550
+ },
+ {
+ "entropy": 0.6600116823613643,
+ "epoch": 1.4021040327293979,
+ "grad_norm": 0.46475180983543396,
+ "learning_rate": 0.00026236025574535523,
+ "loss": 0.6124750518798828,
+ "mean_token_accuracy": 0.8175293403863907,
+ "num_tokens": 1865246.0,
+ "step": 600
+ },
+ {
+ "entropy": 0.6494863349199295,
+ "epoch": 1.5189947399181765,
+ "grad_norm": 0.6132755875587463,
+ "learning_rate": 0.00026150468970660587,
+ "loss": 0.5997534942626953,
+ "mean_token_accuracy": 0.822050162255764,
+ "num_tokens": 2022060.0,
+ "step": 650
+ },
+ {
+ "entropy": 0.6511707258224487,
+ "epoch": 1.635885447106955,
+ "grad_norm": 0.570785641670227,
+ "learning_rate": 0.00026043350045872797,
+ "loss": 0.6059062576293945,
+ "mean_token_accuracy": 0.820236160159111,
+ "num_tokens": 2178909.0,
+ "step": 700
+ },
+ {
+ "entropy": 0.6550412529706955,
+ "epoch": 1.7527761542957334,
+ "grad_norm": 0.5398300290107727,
+ "learning_rate": 0.0002591484690398568,
+ "loss": 0.6031087875366211,
+ "mean_token_accuracy": 0.8206016793847084,
+ "num_tokens": 2329314.0,
+ "step": 750
+ },
+ {
+ "entropy": 0.6509851542115211,
+ "epoch": 1.869666861484512,
+ "grad_norm": 0.6913411617279053,
+ "learning_rate": 0.0002576517320378704,
+ "loss": 0.6023480224609375,
+ "mean_token_accuracy": 0.822216115295887,
+ "num_tokens": 2478559.0,
+ "step": 800
+ },
+ {
+ "entropy": 0.6560930109024048,
+ "epoch": 1.9865575686732906,
+ "grad_norm": 0.5568020939826965,
+ "learning_rate": 0.0002559457780379408,
+ "loss": 0.6031641006469727,
+ "mean_token_accuracy": 0.8205807811021805,
+ "num_tokens": 2631937.0,
+ "step": 850
+ },
+ {
+ "epoch": 2.0,
+ "eval_entropy": 0.794637631866839,
+ "eval_loss": 0.7390629649162292,
+ "eval_mean_token_accuracy": 0.7983863295672776,
+ "eval_num_tokens": 2651286.0,
+ "eval_runtime": 36.7346,
+ "eval_samples_per_second": 33.511,
+ "eval_steps_per_second": 4.192,
+ "step": 856
+ },
+ {
+ "entropy": 0.5709320336730037,
+ "epoch": 2.102863822326125,
+ "grad_norm": 0.4339260458946228,
+ "learning_rate": 0.0002540334434848298,
+ "loss": 0.5152394485473633,
+ "mean_token_accuracy": 0.8391561669919958,
+ "num_tokens": 2795200.0,
+ "step": 900
+ },
+ {
+ "entropy": 0.5768725140392781,
+ "epoch": 2.2197545295149035,
+ "grad_norm": 0.6345565319061279,
+ "learning_rate": 0.00025191790796680617,
+ "loss": 0.5241253662109375,
+ "mean_token_accuracy": 0.8377170553803444,
+ "num_tokens": 2950534.0,
+ "step": 950
+ },
+ {
+ "entropy": 0.570428695231676,
+ "epoch": 2.3366452367036823,
+ "grad_norm": 0.5838360786437988,
+ "learning_rate": 0.00024960268892902876,
+ "loss": 0.5187693405151367,
+ "mean_token_accuracy": 0.8382867723703384,
+ "num_tokens": 3101471.0,
+ "step": 1000
+ },
+ {
+ "entropy": 0.5713468404114246,
+ "epoch": 2.4535359438924607,
+ "grad_norm": 0.5641153454780579,
+ "learning_rate": 0.0002470916358251826,
+ "loss": 0.5234761810302735,
+ "mean_token_accuracy": 0.8382993021607399,
+ "num_tokens": 3258554.0,
+ "step": 1050
+ },
+ {
+ "entropy": 0.5645828676223755,
+ "epoch": 2.570426651081239,
+ "grad_norm": 0.504467785358429,
+ "learning_rate": 0.0002443889237170939,
+ "loss": 0.5156657409667968,
+ "mean_token_accuracy": 0.8391393664479255,
+ "num_tokens": 3416088.0,
+ "step": 1100
+ },
+ {
+ "entropy": 0.5659542319178581,
+ "epoch": 2.6873173582700174,
+ "grad_norm": 0.5225860476493835,
+ "learning_rate": 0.00024149904633296492,
+ "loss": 0.5180046844482422,
+ "mean_token_accuracy": 0.8389606699347496,
+ "num_tokens": 3567251.0,
+ "step": 1150
+ },
+ {
+ "entropy": 0.5713718670606613,
+ "epoch": 2.8042080654587958,
+ "grad_norm": 0.4986768066883087,
+ "learning_rate": 0.0002384268085957702,
+ "loss": 0.524637565612793,
+ "mean_token_accuracy": 0.8381215763092041,
+ "num_tokens": 3719464.0,
+ "step": 1200
+ },
+ {
+ "entropy": 0.5710382495820522,
+ "epoch": 2.9210987726475746,
+ "grad_norm": 0.5548401474952698,
+ "learning_rate": 0.00023517731863423788,
+ "loss": 0.5251939773559571,
+ "mean_token_accuracy": 0.8378310143947602,
+ "num_tokens": 3874423.0,
+ "step": 1250
+ },
+ {
+ "epoch": 3.0,
+ "eval_entropy": 0.6970683218209774,
+ "eval_loss": 0.7369498014450073,
+ "eval_mean_token_accuracy": 0.7919643076209278,
+ "eval_num_tokens": 3976929.0,
+ "eval_runtime": 36.9722,
+ "eval_samples_per_second": 33.295,
+ "eval_steps_per_second": 4.165,
+ "step": 1284
+ },
+ {
+ "entropy": 0.5407733828877684,
+ "epoch": 3.037405026300409,
+ "grad_norm": 0.6430732607841492,
+ "learning_rate": 0.00023175597928969878,
+ "loss": 0.49219757080078125,
+ "mean_token_accuracy": 0.8467060854686564,
+ "num_tokens": 4027609.0,
+ "step": 1300
+ },
+ {
+ "entropy": 0.4786584210395813,
+ "epoch": 3.1542957334891875,
+ "grad_norm": 0.6474543809890747,
+ "learning_rate": 0.00022816847913292434,
+ "loss": 0.4224985122680664,
+ "mean_token_accuracy": 0.8628773537278175,
+ "num_tokens": 4180560.0,
+ "step": 1350
+ },
+ {
+ "entropy": 0.49172899857163427,
+ "epoch": 3.2711864406779663,
+ "grad_norm": 0.549326479434967,
+ "learning_rate": 0.00022442078300588989,
+ "loss": 0.43718509674072265,
+ "mean_token_accuracy": 0.8583108323812485,
+ "num_tokens": 4329954.0,
+ "step": 1400
+ },
+ {
+ "entropy": 0.4829080794751644,
+ "epoch": 3.3880771478667446,
+ "grad_norm": 0.5274888873100281,
+ "learning_rate": 0.00022051912210418922,
+ "loss": 0.4315507125854492,
+ "mean_token_accuracy": 0.8612637943029404,
+ "num_tokens": 4484354.0,
+ "step": 1450
+ },
+ {
+ "entropy": 0.4855460473895073,
+ "epoch": 3.504967855055523,
+ "grad_norm": 0.5846306681632996,
+ "learning_rate": 0.00021646998361658985,
+ "loss": 0.43724521636962893,
+ "mean_token_accuracy": 0.8589270997047425,
+ "num_tokens": 4638919.0,
+ "step": 1500
+ },
+ {
+ "entropy": 0.4905838692188263,
+ "epoch": 3.6218585622443014,
+ "grad_norm": 0.5149123072624207,
+ "learning_rate": 0.00021228009993895495,
+ "loss": 0.44197372436523436,
+ "mean_token_accuracy": 0.8584243017435074,
+ "num_tokens": 4795540.0,
+ "step": 1550
+ },
+ {
+ "entropy": 0.4994237454235554,
+ "epoch": 3.73874926943308,
+ "grad_norm": 0.5417225956916809,
+ "learning_rate": 0.0002079564374804665,
+ "loss": 0.4491108322143555,
+ "mean_token_accuracy": 0.8556393891572952,
+ "num_tokens": 4951177.0,
+ "step": 1600
+ },
+ {
+ "entropy": 0.49903531357645986,
+ "epoch": 3.8556399766218585,
+ "grad_norm": 0.5086734890937805,
+ "learning_rate": 0.00020350618508076015,
+ "loss": 0.4518918228149414,
+ "mean_token_accuracy": 0.8549707370996475,
+ "num_tokens": 5104814.0,
+ "step": 1650
+ },
+ {
+ "entropy": 0.4859523306787014,
+ "epoch": 3.972530683810637,
+ "grad_norm": 0.5446717739105225,
+ "learning_rate": 0.00019893674205723094,
+ "loss": 0.4410313415527344,
+ "mean_token_accuracy": 0.8587654489278793,
+ "num_tokens": 5265185.0,
+ "step": 1700
+ },
+ {
+ "epoch": 4.0,
+ "eval_entropy": 0.6577693150415049,
+ "eval_loss": 0.7380754947662354,
+ "eval_mean_token_accuracy": 0.8039549565934515,
+ "eval_num_tokens": 5302572.0,
+ "eval_runtime": 36.5757,
+ "eval_samples_per_second": 33.656,
+ "eval_steps_per_second": 4.21,
+ "step": 1712
+ },
+ {
+ "entropy": 0.41945636961328325,
+ "epoch": 4.088836937463472,
+ "grad_norm": 0.5925881266593933,
+ "learning_rate": 0.00019425570590238332,
+ "loss": 0.3575031280517578,
+ "mean_token_accuracy": 0.8812823918596584,
+ "num_tokens": 5419002.0,
+ "step": 1750
+ },
+ {
+ "entropy": 0.40106018364429474,
+ "epoch": 4.20572764465225,
+ "grad_norm": 0.7094829082489014,
+ "learning_rate": 0.0001894708596516806,
+ "loss": 0.3372145080566406,
+ "mean_token_accuracy": 0.8856011089682579,
+ "num_tokens": 5572969.0,
+ "step": 1800
+ },
+ {
+ "entropy": 0.4121378688514233,
+ "epoch": 4.322618351841029,
+ "grad_norm": 0.6077721118927002,
+ "learning_rate": 0.00018459015894289695,
+ "loss": 0.34956165313720705,
+ "mean_token_accuracy": 0.8818695649504662,
+ "num_tokens": 5733805.0,
+ "step": 1850
+ },
+ {
+ "entropy": 0.40683484479784965,
+ "epoch": 4.439509059029807,
+ "grad_norm": 0.5614883303642273,
+ "learning_rate": 0.00017962171878848807,
+ "loss": 0.34749691009521483,
+ "mean_token_accuracy": 0.882930768430233,
+ "num_tokens": 5888694.0,
+ "step": 1900
+ },
+ {
+ "entropy": 0.4100585526227951,
+ "epoch": 4.556399766218585,
+ "grad_norm": 0.5506443977355957,
+ "learning_rate": 0.00017457380008297383,
+ "loss": 0.3501173782348633,
+ "mean_token_accuracy": 0.8827524068951607,
+ "num_tokens": 6042655.0,
+ "step": 1950
+ },
+ {
+ "entropy": 0.41173958137631417,
+ "epoch": 4.673290473407365,
+ "grad_norm": 0.5409566164016724,
+ "learning_rate": 0.00016945479586776672,
+ "loss": 0.3541391754150391,
+ "mean_token_accuracy": 0.8808356913924217,
+ "num_tokens": 6203980.0,
+ "step": 2000
+ },
+ {
+ "entropy": 0.4158783262968063,
+ "epoch": 4.790181180596143,
+ "grad_norm": 0.572618842124939,
+ "learning_rate": 0.00016427321737628295,
+ "loss": 0.3536404800415039,
+ "mean_token_accuracy": 0.880456589460373,
+ "num_tokens": 6354604.0,
+ "step": 2050
+ },
+ {
+ "entropy": 0.408166281580925,
+ "epoch": 4.907071887784921,
+ "grad_norm": 0.5987762808799744,
+ "learning_rate": 0.00015903767988253897,
+ "loss": 0.3518599319458008,
+ "mean_token_accuracy": 0.8815453574061394,
+ "num_tokens": 6507484.0,
+ "step": 2100
+ },
+ {
+ "epoch": 5.0,
+ "eval_entropy": 0.6047254425364655,
+ "eval_loss": 0.7816920876502991,
+ "eval_mean_token_accuracy": 0.7987685733801358,
+ "eval_num_tokens": 6628215.0,
+ "eval_runtime": 36.9374,
+ "eval_samples_per_second": 33.327,
+ "eval_steps_per_second": 4.169,
+ "step": 2140
+ },
+ {
+ "entropy": 0.38985791001188097,
+ "epoch": 5.023378141437756,
+ "grad_norm": 0.7454882860183716,
+ "learning_rate": 0.00015375688837676246,
+ "loss": 0.32916854858398437,
+ "mean_token_accuracy": 0.8886270373310875,
+ "num_tokens": 6662015.0,
+ "step": 2150
+ },
+ {
+ "entropy": 0.30170531339943407,
+ "epoch": 5.140268848626534,
+ "grad_norm": 0.5329811573028564,
+ "learning_rate": 0.0001484396230918344,
+ "loss": 0.23112209320068358,
+ "mean_token_accuracy": 0.9198204749822616,
+ "num_tokens": 6817989.0,
+ "step": 2200
+ },
+ {
+ "entropy": 0.3075008810311556,
+ "epoch": 5.257159555815313,
+ "grad_norm": 0.6896093487739563,
+ "learning_rate": 0.00014309472490462715,
+ "loss": 0.2416652488708496,
+ "mean_token_accuracy": 0.9158366405963898,
+ "num_tokens": 6977021.0,
+ "step": 2250
+ },
+ {
+ "entropy": 0.31153449676930905,
+ "epoch": 5.374050263004091,
+ "grad_norm": 0.6930833458900452,
+ "learning_rate": 0.00013773108063651161,
+ "loss": 0.24753889083862304,
+ "mean_token_accuracy": 0.9136614826321602,
+ "num_tokens": 7133360.0,
+ "step": 2300
+ },
+ {
+ "entropy": 0.3154346436262131,
+ "epoch": 5.490940970192869,
+ "grad_norm": 0.6476582884788513,
+ "learning_rate": 0.0001323576082774734,
+ "loss": 0.24872392654418946,
+ "mean_token_accuracy": 0.9135914054512978,
+ "num_tokens": 7282372.0,
+ "step": 2350
+ },
+ {
+ "entropy": 0.313641587048769,
+ "epoch": 5.607831677381649,
+ "grad_norm": 0.6953482627868652,
+ "learning_rate": 0.0001269832421584059,
+ "loss": 0.24919464111328124,
+ "mean_token_accuracy": 0.9131662029027939,
+ "num_tokens": 7435766.0,
+ "step": 2400
+ },
+ {
+ "entropy": 0.3163894522190094,
+ "epoch": 5.724722384570427,
+ "grad_norm": 0.6673940420150757,
+ "learning_rate": 0.00012161691809623384,
+ "loss": 0.25111726760864256,
+ "mean_token_accuracy": 0.913602537214756,
+ "num_tokens": 7578408.0,
+ "step": 2450
+ },
+ {
+ "entropy": 0.3183368914574385,
+ "epoch": 5.841613091759205,
+ "grad_norm": 0.7174487709999084,
+ "learning_rate": 0.00011626755853656576,
+ "loss": 0.2518378448486328,
+ "mean_token_accuracy": 0.912000589966774,
+ "num_tokens": 7734599.0,
+ "step": 2500
+ },
+ {
+ "entropy": 0.3077564483135939,
+ "epoch": 5.958503798947984,
+ "grad_norm": 0.6662200689315796,
+ "learning_rate": 0.00011094405771857891,
+ "loss": 0.24354087829589843,
+ "mean_token_accuracy": 0.9140791621804237,
+ "num_tokens": 7898028.0,
+ "step": 2550
+ },
+ {
+ "epoch": 6.0,
+ "eval_entropy": 0.5403795917699863,
+ "eval_loss": 0.8572548627853394,
+ "eval_mean_token_accuracy": 0.796143418395674,
+ "eval_num_tokens": 7953858.0,
+ "eval_runtime": 36.7176,
+ "eval_samples_per_second": 33.526,
+ "eval_steps_per_second": 4.194,
+ "step": 2568
+ },
+ {
+ "entropy": 0.25110363144071857,
+ "epoch": 6.074810052600818,
+ "grad_norm": 0.6466763019561768,
+ "learning_rate": 0.0001056552668868023,
+ "loss": 0.1814822769165039,
+ "mean_token_accuracy": 0.9370300626035911,
+ "num_tokens": 8052646.0,
+ "step": 2600
+ },
+ {
+ "entropy": 0.2063255251944065,
+ "epoch": 6.1917007597895966,
+ "grad_norm": 0.6950973272323608,
+ "learning_rate": 0.0001004099795743853,
+ "loss": 0.13960565567016603,
+ "mean_token_accuracy": 0.9511301657557487,
+ "num_tokens": 8212323.0,
+ "step": 2650
+ },
+ {
+ "entropy": 0.20775293290615082,
+ "epoch": 6.308591466978375,
+ "grad_norm": 0.660478413105011,
+ "learning_rate": 9.52169169823223e-05,
+ "loss": 0.14319146156311036,
+ "mean_token_accuracy": 0.9500676044821739,
+ "num_tokens": 8366715.0,
+ "step": 2700
+ },
+ {
+ "entropy": 0.20977566957473756,
+ "epoch": 6.425482174167154,
+ "grad_norm": 0.6141477823257446,
+ "learning_rate": 9.00847134789411e-05,
+ "loss": 0.14525352478027342,
+ "mean_token_accuracy": 0.9490846493840217,
+ "num_tokens": 8515955.0,
+ "step": 2750
+ },
+ {
+ "entropy": 0.2143413795530796,
+ "epoch": 6.5423728813559325,
+ "grad_norm": 0.612184464931488,
+ "learning_rate": 8.502190224376661e-05,
+ "loss": 0.14751774787902833,
+ "mean_token_accuracy": 0.9478235357999801,
+ "num_tokens": 8672361.0,
+ "step": 2800
+ },
+ {
+ "entropy": 0.21027401253581046,
+ "epoch": 6.659263588544711,
+ "grad_norm": 0.644019365310669,
+ "learning_rate": 8.003690107962787e-05,
+ "loss": 0.1495517921447754,
+ "mean_token_accuracy": 0.9470956793427467,
+ "num_tokens": 8827819.0,
+ "step": 2850
+ },
+ {
+ "entropy": 0.2109992092102766,
+ "epoch": 6.776154295733489,
+ "grad_norm": 0.5456500053405762,
+ "learning_rate": 7.513799841659927e-05,
+ "loss": 0.14893126487731934,
+ "mean_token_accuracy": 0.9479587042331695,
+ "num_tokens": 8979535.0,
+ "step": 2900
+ },
+ {
+ "entropy": 0.20949276342988013,
+ "epoch": 6.893045002922268,
+ "grad_norm": 0.6803008913993835,
+ "learning_rate": 7.033333953104615e-05,
+ "loss": 0.1461741542816162,
+ "mean_token_accuracy": 0.9480843073129654,
+ "num_tokens": 9134134.0,
+ "step": 2950
+ },
+ {
+ "epoch": 7.0,
+ "eval_entropy": 0.4759114793949313,
+ "eval_loss": 0.9827883839607239,
+ "eval_mean_token_accuracy": 0.7945680889216337,
+ "eval_num_tokens": 9279501.0,
+ "eval_runtime": 36.4188,
+ "eval_samples_per_second": 33.801,
+ "eval_steps_per_second": 4.229,
+ "step": 2996
+ }
+ ],
+ "logging_steps": 50,
+ "max_steps": 4280,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 10,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.4340983113207296e+17,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3424/README.md b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3424/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3424/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3424/adapter_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3424/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9e45544de727016be134f25e606b7b2b87c67549
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3424/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.050044417519510255,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "v_proj",
+ "o_proj",
+ "q_proj",
+ "gate_proj",
+ "k_proj",
+ "down_proj",
+ "up_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3424/chat_template.jinja b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3424/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3424/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3424/tokenizer_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3424/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3424/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3424/trainer_state.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3424/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6f3bf7a177ee42ac40a722b38f4f97628a909364
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3424/trainer_state.json
@@ -0,0 +1,802 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 8.0,
+ "eval_steps": 500,
+ "global_step": 3424,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 1.8448095142841339,
+ "epoch": 0.11689070718877849,
+ "grad_norm": 1.4738068580627441,
+ "learning_rate": 3.0183098859435753e-05,
+ "loss": 1.6948321533203126,
+ "mean_token_accuracy": 0.6315939916670322,
+ "num_tokens": 153689.0,
+ "step": 50
+ },
+ {
+ "entropy": 0.9560942512750625,
+ "epoch": 0.23378141437755698,
+ "grad_norm": 1.378090739250183,
+ "learning_rate": 6.098217932824775e-05,
+ "loss": 0.8888477325439453,
+ "mean_token_accuracy": 0.7586692428588867,
+ "num_tokens": 310548.0,
+ "step": 100
+ },
+ {
+ "entropy": 0.848925698697567,
+ "epoch": 0.3506721215663355,
+ "grad_norm": 0.959621250629425,
+ "learning_rate": 9.178125979705974e-05,
+ "loss": 0.7850731658935547,
+ "mean_token_accuracy": 0.7774798944592476,
+ "num_tokens": 468560.0,
+ "step": 150
+ },
+ {
+ "entropy": 0.811761694252491,
+ "epoch": 0.46756282875511396,
+ "grad_norm": 0.8867514133453369,
+ "learning_rate": 0.00012258034026587175,
+ "loss": 0.7446884155273438,
+ "mean_token_accuracy": 0.7892421445250511,
+ "num_tokens": 619552.0,
+ "step": 200
+ },
+ {
+ "entropy": 0.7709847274422645,
+ "epoch": 0.5844535359438925,
+ "grad_norm": 0.5822142362594604,
+ "learning_rate": 0.00015337942073468373,
+ "loss": 0.7130480194091797,
+ "mean_token_accuracy": 0.7964023387432099,
+ "num_tokens": 771972.0,
+ "step": 250
+ },
+ {
+ "entropy": 0.7401632118225098,
+ "epoch": 0.701344243132671,
+ "grad_norm": 0.7557511925697327,
+ "learning_rate": 0.00018417850120349574,
+ "loss": 0.6857429504394531,
+ "mean_token_accuracy": 0.8041065946221352,
+ "num_tokens": 928088.0,
+ "step": 300
+ },
+ {
+ "entropy": 0.7295274111628532,
+ "epoch": 0.8182349503214494,
+ "grad_norm": 0.6769705414772034,
+ "learning_rate": 0.00021497758167230772,
+ "loss": 0.6801216125488281,
+ "mean_token_accuracy": 0.8037989619374275,
+ "num_tokens": 1086836.0,
+ "step": 350
+ },
+ {
+ "entropy": 0.710282272696495,
+ "epoch": 0.9351256575102279,
+ "grad_norm": 0.832009494304657,
+ "learning_rate": 0.00024577666214111967,
+ "loss": 0.6587177276611328,
+ "mean_token_accuracy": 0.8084024578332901,
+ "num_tokens": 1241227.0,
+ "step": 400
+ },
+ {
+ "epoch": 1.0,
+ "eval_entropy": 0.848237401672772,
+ "eval_loss": 0.7832363247871399,
+ "eval_mean_token_accuracy": 0.7817642971292719,
+ "eval_num_tokens": 1325643.0,
+ "eval_runtime": 36.8595,
+ "eval_samples_per_second": 33.397,
+ "eval_steps_per_second": 4.178,
+ "step": 428
+ },
+ {
+ "entropy": 0.7039205249230466,
+ "epoch": 1.0514319111630626,
+ "grad_norm": 0.6213153600692749,
+ "learning_rate": 0.00026362079548206905,
+ "loss": 0.6527813720703125,
+ "mean_token_accuracy": 0.8083000521564004,
+ "num_tokens": 1393922.0,
+ "step": 450
+ },
+ {
+ "entropy": 0.6749556151032448,
+ "epoch": 1.1683226183518411,
+ "grad_norm": 0.6444424390792847,
+ "learning_rate": 0.0002634191889632654,
+ "loss": 0.627038345336914,
+ "mean_token_accuracy": 0.8141268998384475,
+ "num_tokens": 1551442.0,
+ "step": 500
+ },
+ {
+ "entropy": 0.6727319830656051,
+ "epoch": 1.2852133255406195,
+ "grad_norm": 0.6360110640525818,
+ "learning_rate": 0.00026299877604786883,
+ "loss": 0.6232248687744141,
+ "mean_token_accuracy": 0.8153417313098907,
+ "num_tokens": 1706697.0,
+ "step": 550
+ },
+ {
+ "entropy": 0.6600116823613643,
+ "epoch": 1.4021040327293979,
+ "grad_norm": 0.46475180983543396,
+ "learning_rate": 0.00026236025574535523,
+ "loss": 0.6124750518798828,
+ "mean_token_accuracy": 0.8175293403863907,
+ "num_tokens": 1865246.0,
+ "step": 600
+ },
+ {
+ "entropy": 0.6494863349199295,
+ "epoch": 1.5189947399181765,
+ "grad_norm": 0.6132755875587463,
+ "learning_rate": 0.00026150468970660587,
+ "loss": 0.5997534942626953,
+ "mean_token_accuracy": 0.822050162255764,
+ "num_tokens": 2022060.0,
+ "step": 650
+ },
+ {
+ "entropy": 0.6511707258224487,
+ "epoch": 1.635885447106955,
+ "grad_norm": 0.570785641670227,
+ "learning_rate": 0.00026043350045872797,
+ "loss": 0.6059062576293945,
+ "mean_token_accuracy": 0.820236160159111,
+ "num_tokens": 2178909.0,
+ "step": 700
+ },
+ {
+ "entropy": 0.6550412529706955,
+ "epoch": 1.7527761542957334,
+ "grad_norm": 0.5398300290107727,
+ "learning_rate": 0.0002591484690398568,
+ "loss": 0.6031087875366211,
+ "mean_token_accuracy": 0.8206016793847084,
+ "num_tokens": 2329314.0,
+ "step": 750
+ },
+ {
+ "entropy": 0.6509851542115211,
+ "epoch": 1.869666861484512,
+ "grad_norm": 0.6913411617279053,
+ "learning_rate": 0.0002576517320378704,
+ "loss": 0.6023480224609375,
+ "mean_token_accuracy": 0.822216115295887,
+ "num_tokens": 2478559.0,
+ "step": 800
+ },
+ {
+ "entropy": 0.6560930109024048,
+ "epoch": 1.9865575686732906,
+ "grad_norm": 0.5568020939826965,
+ "learning_rate": 0.0002559457780379408,
+ "loss": 0.6031641006469727,
+ "mean_token_accuracy": 0.8205807811021805,
+ "num_tokens": 2631937.0,
+ "step": 850
+ },
+ {
+ "epoch": 2.0,
+ "eval_entropy": 0.794637631866839,
+ "eval_loss": 0.7390629649162292,
+ "eval_mean_token_accuracy": 0.7983863295672776,
+ "eval_num_tokens": 2651286.0,
+ "eval_runtime": 36.7346,
+ "eval_samples_per_second": 33.511,
+ "eval_steps_per_second": 4.192,
+ "step": 856
+ },
+ {
+ "entropy": 0.5709320336730037,
+ "epoch": 2.102863822326125,
+ "grad_norm": 0.4339260458946228,
+ "learning_rate": 0.0002540334434848298,
+ "loss": 0.5152394485473633,
+ "mean_token_accuracy": 0.8391561669919958,
+ "num_tokens": 2795200.0,
+ "step": 900
+ },
+ {
+ "entropy": 0.5768725140392781,
+ "epoch": 2.2197545295149035,
+ "grad_norm": 0.6345565319061279,
+ "learning_rate": 0.00025191790796680617,
+ "loss": 0.5241253662109375,
+ "mean_token_accuracy": 0.8377170553803444,
+ "num_tokens": 2950534.0,
+ "step": 950
+ },
+ {
+ "entropy": 0.570428695231676,
+ "epoch": 2.3366452367036823,
+ "grad_norm": 0.5838360786437988,
+ "learning_rate": 0.00024960268892902876,
+ "loss": 0.5187693405151367,
+ "mean_token_accuracy": 0.8382867723703384,
+ "num_tokens": 3101471.0,
+ "step": 1000
+ },
+ {
+ "entropy": 0.5713468404114246,
+ "epoch": 2.4535359438924607,
+ "grad_norm": 0.5641153454780579,
+ "learning_rate": 0.0002470916358251826,
+ "loss": 0.5234761810302735,
+ "mean_token_accuracy": 0.8382993021607399,
+ "num_tokens": 3258554.0,
+ "step": 1050
+ },
+ {
+ "entropy": 0.5645828676223755,
+ "epoch": 2.570426651081239,
+ "grad_norm": 0.504467785358429,
+ "learning_rate": 0.0002443889237170939,
+ "loss": 0.5156657409667968,
+ "mean_token_accuracy": 0.8391393664479255,
+ "num_tokens": 3416088.0,
+ "step": 1100
+ },
+ {
+ "entropy": 0.5659542319178581,
+ "epoch": 2.6873173582700174,
+ "grad_norm": 0.5225860476493835,
+ "learning_rate": 0.00024149904633296492,
+ "loss": 0.5180046844482422,
+ "mean_token_accuracy": 0.8389606699347496,
+ "num_tokens": 3567251.0,
+ "step": 1150
+ },
+ {
+ "entropy": 0.5713718670606613,
+ "epoch": 2.8042080654587958,
+ "grad_norm": 0.4986768066883087,
+ "learning_rate": 0.0002384268085957702,
+ "loss": 0.524637565612793,
+ "mean_token_accuracy": 0.8381215763092041,
+ "num_tokens": 3719464.0,
+ "step": 1200
+ },
+ {
+ "entropy": 0.5710382495820522,
+ "epoch": 2.9210987726475746,
+ "grad_norm": 0.5548401474952698,
+ "learning_rate": 0.00023517731863423788,
+ "loss": 0.5251939773559571,
+ "mean_token_accuracy": 0.8378310143947602,
+ "num_tokens": 3874423.0,
+ "step": 1250
+ },
+ {
+ "epoch": 3.0,
+ "eval_entropy": 0.6970683218209774,
+ "eval_loss": 0.7369498014450073,
+ "eval_mean_token_accuracy": 0.7919643076209278,
+ "eval_num_tokens": 3976929.0,
+ "eval_runtime": 36.9722,
+ "eval_samples_per_second": 33.295,
+ "eval_steps_per_second": 4.165,
+ "step": 1284
+ },
+ {
+ "entropy": 0.5407733828877684,
+ "epoch": 3.037405026300409,
+ "grad_norm": 0.6430732607841492,
+ "learning_rate": 0.00023175597928969878,
+ "loss": 0.49219757080078125,
+ "mean_token_accuracy": 0.8467060854686564,
+ "num_tokens": 4027609.0,
+ "step": 1300
+ },
+ {
+ "entropy": 0.4786584210395813,
+ "epoch": 3.1542957334891875,
+ "grad_norm": 0.6474543809890747,
+ "learning_rate": 0.00022816847913292434,
+ "loss": 0.4224985122680664,
+ "mean_token_accuracy": 0.8628773537278175,
+ "num_tokens": 4180560.0,
+ "step": 1350
+ },
+ {
+ "entropy": 0.49172899857163427,
+ "epoch": 3.2711864406779663,
+ "grad_norm": 0.549326479434967,
+ "learning_rate": 0.00022442078300588989,
+ "loss": 0.43718509674072265,
+ "mean_token_accuracy": 0.8583108323812485,
+ "num_tokens": 4329954.0,
+ "step": 1400
+ },
+ {
+ "entropy": 0.4829080794751644,
+ "epoch": 3.3880771478667446,
+ "grad_norm": 0.5274888873100281,
+ "learning_rate": 0.00022051912210418922,
+ "loss": 0.4315507125854492,
+ "mean_token_accuracy": 0.8612637943029404,
+ "num_tokens": 4484354.0,
+ "step": 1450
+ },
+ {
+ "entropy": 0.4855460473895073,
+ "epoch": 3.504967855055523,
+ "grad_norm": 0.5846306681632996,
+ "learning_rate": 0.00021646998361658985,
+ "loss": 0.43724521636962893,
+ "mean_token_accuracy": 0.8589270997047425,
+ "num_tokens": 4638919.0,
+ "step": 1500
+ },
+ {
+ "entropy": 0.4905838692188263,
+ "epoch": 3.6218585622443014,
+ "grad_norm": 0.5149123072624207,
+ "learning_rate": 0.00021228009993895495,
+ "loss": 0.44197372436523436,
+ "mean_token_accuracy": 0.8584243017435074,
+ "num_tokens": 4795540.0,
+ "step": 1550
+ },
+ {
+ "entropy": 0.4994237454235554,
+ "epoch": 3.73874926943308,
+ "grad_norm": 0.5417225956916809,
+ "learning_rate": 0.0002079564374804665,
+ "loss": 0.4491108322143555,
+ "mean_token_accuracy": 0.8556393891572952,
+ "num_tokens": 4951177.0,
+ "step": 1600
+ },
+ {
+ "entropy": 0.49903531357645986,
+ "epoch": 3.8556399766218585,
+ "grad_norm": 0.5086734890937805,
+ "learning_rate": 0.00020350618508076015,
+ "loss": 0.4518918228149414,
+ "mean_token_accuracy": 0.8549707370996475,
+ "num_tokens": 5104814.0,
+ "step": 1650
+ },
+ {
+ "entropy": 0.4859523306787014,
+ "epoch": 3.972530683810637,
+ "grad_norm": 0.5446717739105225,
+ "learning_rate": 0.00019893674205723094,
+ "loss": 0.4410313415527344,
+ "mean_token_accuracy": 0.8587654489278793,
+ "num_tokens": 5265185.0,
+ "step": 1700
+ },
+ {
+ "epoch": 4.0,
+ "eval_entropy": 0.6577693150415049,
+ "eval_loss": 0.7380754947662354,
+ "eval_mean_token_accuracy": 0.8039549565934515,
+ "eval_num_tokens": 5302572.0,
+ "eval_runtime": 36.5757,
+ "eval_samples_per_second": 33.656,
+ "eval_steps_per_second": 4.21,
+ "step": 1712
+ },
+ {
+ "entropy": 0.41945636961328325,
+ "epoch": 4.088836937463472,
+ "grad_norm": 0.5925881266593933,
+ "learning_rate": 0.00019425570590238332,
+ "loss": 0.3575031280517578,
+ "mean_token_accuracy": 0.8812823918596584,
+ "num_tokens": 5419002.0,
+ "step": 1750
+ },
+ {
+ "entropy": 0.40106018364429474,
+ "epoch": 4.20572764465225,
+ "grad_norm": 0.7094829082489014,
+ "learning_rate": 0.0001894708596516806,
+ "loss": 0.3372145080566406,
+ "mean_token_accuracy": 0.8856011089682579,
+ "num_tokens": 5572969.0,
+ "step": 1800
+ },
+ {
+ "entropy": 0.4121378688514233,
+ "epoch": 4.322618351841029,
+ "grad_norm": 0.6077721118927002,
+ "learning_rate": 0.00018459015894289695,
+ "loss": 0.34956165313720705,
+ "mean_token_accuracy": 0.8818695649504662,
+ "num_tokens": 5733805.0,
+ "step": 1850
+ },
+ {
+ "entropy": 0.40683484479784965,
+ "epoch": 4.439509059029807,
+ "grad_norm": 0.5614883303642273,
+ "learning_rate": 0.00017962171878848807,
+ "loss": 0.34749691009521483,
+ "mean_token_accuracy": 0.882930768430233,
+ "num_tokens": 5888694.0,
+ "step": 1900
+ },
+ {
+ "entropy": 0.4100585526227951,
+ "epoch": 4.556399766218585,
+ "grad_norm": 0.5506443977355957,
+ "learning_rate": 0.00017457380008297383,
+ "loss": 0.3501173782348633,
+ "mean_token_accuracy": 0.8827524068951607,
+ "num_tokens": 6042655.0,
+ "step": 1950
+ },
+ {
+ "entropy": 0.41173958137631417,
+ "epoch": 4.673290473407365,
+ "grad_norm": 0.5409566164016724,
+ "learning_rate": 0.00016945479586776672,
+ "loss": 0.3541391754150391,
+ "mean_token_accuracy": 0.8808356913924217,
+ "num_tokens": 6203980.0,
+ "step": 2000
+ },
+ {
+ "entropy": 0.4158783262968063,
+ "epoch": 4.790181180596143,
+ "grad_norm": 0.572618842124939,
+ "learning_rate": 0.00016427321737628295,
+ "loss": 0.3536404800415039,
+ "mean_token_accuracy": 0.880456589460373,
+ "num_tokens": 6354604.0,
+ "step": 2050
+ },
+ {
+ "entropy": 0.408166281580925,
+ "epoch": 4.907071887784921,
+ "grad_norm": 0.5987762808799744,
+ "learning_rate": 0.00015903767988253897,
+ "loss": 0.3518599319458008,
+ "mean_token_accuracy": 0.8815453574061394,
+ "num_tokens": 6507484.0,
+ "step": 2100
+ },
+ {
+ "epoch": 5.0,
+ "eval_entropy": 0.6047254425364655,
+ "eval_loss": 0.7816920876502991,
+ "eval_mean_token_accuracy": 0.7987685733801358,
+ "eval_num_tokens": 6628215.0,
+ "eval_runtime": 36.9374,
+ "eval_samples_per_second": 33.327,
+ "eval_steps_per_second": 4.169,
+ "step": 2140
+ },
+ {
+ "entropy": 0.38985791001188097,
+ "epoch": 5.023378141437756,
+ "grad_norm": 0.7454882860183716,
+ "learning_rate": 0.00015375688837676246,
+ "loss": 0.32916854858398437,
+ "mean_token_accuracy": 0.8886270373310875,
+ "num_tokens": 6662015.0,
+ "step": 2150
+ },
+ {
+ "entropy": 0.30170531339943407,
+ "epoch": 5.140268848626534,
+ "grad_norm": 0.5329811573028564,
+ "learning_rate": 0.0001484396230918344,
+ "loss": 0.23112209320068358,
+ "mean_token_accuracy": 0.9198204749822616,
+ "num_tokens": 6817989.0,
+ "step": 2200
+ },
+ {
+ "entropy": 0.3075008810311556,
+ "epoch": 5.257159555815313,
+ "grad_norm": 0.6896093487739563,
+ "learning_rate": 0.00014309472490462715,
+ "loss": 0.2416652488708496,
+ "mean_token_accuracy": 0.9158366405963898,
+ "num_tokens": 6977021.0,
+ "step": 2250
+ },
+ {
+ "entropy": 0.31153449676930905,
+ "epoch": 5.374050263004091,
+ "grad_norm": 0.6930833458900452,
+ "learning_rate": 0.00013773108063651161,
+ "loss": 0.24753889083862304,
+ "mean_token_accuracy": 0.9136614826321602,
+ "num_tokens": 7133360.0,
+ "step": 2300
+ },
+ {
+ "entropy": 0.3154346436262131,
+ "epoch": 5.490940970192869,
+ "grad_norm": 0.6476582884788513,
+ "learning_rate": 0.0001323576082774734,
+ "loss": 0.24872392654418946,
+ "mean_token_accuracy": 0.9135914054512978,
+ "num_tokens": 7282372.0,
+ "step": 2350
+ },
+ {
+ "entropy": 0.313641587048769,
+ "epoch": 5.607831677381649,
+ "grad_norm": 0.6953482627868652,
+ "learning_rate": 0.0001269832421584059,
+ "loss": 0.24919464111328124,
+ "mean_token_accuracy": 0.9131662029027939,
+ "num_tokens": 7435766.0,
+ "step": 2400
+ },
+ {
+ "entropy": 0.3163894522190094,
+ "epoch": 5.724722384570427,
+ "grad_norm": 0.6673940420150757,
+ "learning_rate": 0.00012161691809623384,
+ "loss": 0.25111726760864256,
+ "mean_token_accuracy": 0.913602537214756,
+ "num_tokens": 7578408.0,
+ "step": 2450
+ },
+ {
+ "entropy": 0.3183368914574385,
+ "epoch": 5.841613091759205,
+ "grad_norm": 0.7174487709999084,
+ "learning_rate": 0.00011626755853656576,
+ "loss": 0.2518378448486328,
+ "mean_token_accuracy": 0.912000589966774,
+ "num_tokens": 7734599.0,
+ "step": 2500
+ },
+ {
+ "entropy": 0.3077564483135939,
+ "epoch": 5.958503798947984,
+ "grad_norm": 0.6662200689315796,
+ "learning_rate": 0.00011094405771857891,
+ "loss": 0.24354087829589843,
+ "mean_token_accuracy": 0.9140791621804237,
+ "num_tokens": 7898028.0,
+ "step": 2550
+ },
+ {
+ "epoch": 6.0,
+ "eval_entropy": 0.5403795917699863,
+ "eval_loss": 0.8572548627853394,
+ "eval_mean_token_accuracy": 0.796143418395674,
+ "eval_num_tokens": 7953858.0,
+ "eval_runtime": 36.7176,
+ "eval_samples_per_second": 33.526,
+ "eval_steps_per_second": 4.194,
+ "step": 2568
+ },
+ {
+ "entropy": 0.25110363144071857,
+ "epoch": 6.074810052600818,
+ "grad_norm": 0.6466763019561768,
+ "learning_rate": 0.0001056552668868023,
+ "loss": 0.1814822769165039,
+ "mean_token_accuracy": 0.9370300626035911,
+ "num_tokens": 8052646.0,
+ "step": 2600
+ },
+ {
+ "entropy": 0.2063255251944065,
+ "epoch": 6.1917007597895966,
+ "grad_norm": 0.6950973272323608,
+ "learning_rate": 0.0001004099795743853,
+ "loss": 0.13960565567016603,
+ "mean_token_accuracy": 0.9511301657557487,
+ "num_tokens": 8212323.0,
+ "step": 2650
+ },
+ {
+ "entropy": 0.20775293290615082,
+ "epoch": 6.308591466978375,
+ "grad_norm": 0.660478413105011,
+ "learning_rate": 9.52169169823223e-05,
+ "loss": 0.14319146156311036,
+ "mean_token_accuracy": 0.9500676044821739,
+ "num_tokens": 8366715.0,
+ "step": 2700
+ },
+ {
+ "entropy": 0.20977566957473756,
+ "epoch": 6.425482174167154,
+ "grad_norm": 0.6141477823257446,
+ "learning_rate": 9.00847134789411e-05,
+ "loss": 0.14525352478027342,
+ "mean_token_accuracy": 0.9490846493840217,
+ "num_tokens": 8515955.0,
+ "step": 2750
+ },
+ {
+ "entropy": 0.2143413795530796,
+ "epoch": 6.5423728813559325,
+ "grad_norm": 0.612184464931488,
+ "learning_rate": 8.502190224376661e-05,
+ "loss": 0.14751774787902833,
+ "mean_token_accuracy": 0.9478235357999801,
+ "num_tokens": 8672361.0,
+ "step": 2800
+ },
+ {
+ "entropy": 0.21027401253581046,
+ "epoch": 6.659263588544711,
+ "grad_norm": 0.644019365310669,
+ "learning_rate": 8.003690107962787e-05,
+ "loss": 0.1495517921447754,
+ "mean_token_accuracy": 0.9470956793427467,
+ "num_tokens": 8827819.0,
+ "step": 2850
+ },
+ {
+ "entropy": 0.2109992092102766,
+ "epoch": 6.776154295733489,
+ "grad_norm": 0.5456500053405762,
+ "learning_rate": 7.513799841659927e-05,
+ "loss": 0.14893126487731934,
+ "mean_token_accuracy": 0.9479587042331695,
+ "num_tokens": 8979535.0,
+ "step": 2900
+ },
+ {
+ "entropy": 0.20949276342988013,
+ "epoch": 6.893045002922268,
+ "grad_norm": 0.6803008913993835,
+ "learning_rate": 7.033333953104615e-05,
+ "loss": 0.1461741542816162,
+ "mean_token_accuracy": 0.9480843073129654,
+ "num_tokens": 9134134.0,
+ "step": 2950
+ },
+ {
+ "epoch": 7.0,
+ "eval_entropy": 0.4759114793949313,
+ "eval_loss": 0.9827883839607239,
+ "eval_mean_token_accuracy": 0.7945680889216337,
+ "eval_num_tokens": 9279501.0,
+ "eval_runtime": 36.4188,
+ "eval_samples_per_second": 33.801,
+ "eval_steps_per_second": 4.229,
+ "step": 2996
+ },
+ {
+ "entropy": 0.20382985060838,
+ "epoch": 7.009351256575102,
+ "grad_norm": 0.42079976201057434,
+ "learning_rate": 6.563091300268857e-05,
+ "loss": 0.13976585388183593,
+ "mean_token_accuracy": 0.9511888485458029,
+ "num_tokens": 9290416.0,
+ "step": 3000
+ },
+ {
+ "entropy": 0.14199006125330926,
+ "epoch": 7.1262419637638805,
+ "grad_norm": 0.6262232065200806,
+ "learning_rate": 6.103853743220019e-05,
+ "loss": 0.08045654296875,
+ "mean_token_accuracy": 0.9727814987301826,
+ "num_tokens": 9444705.0,
+ "step": 3050
+ },
+ {
+ "entropy": 0.13681559398770332,
+ "epoch": 7.243132670952659,
+ "grad_norm": 0.6460694074630737,
+ "learning_rate": 5.6563848441427006e-05,
+ "loss": 0.0795314359664917,
+ "mean_token_accuracy": 0.973130469918251,
+ "num_tokens": 9601476.0,
+ "step": 3100
+ },
+ {
+ "entropy": 0.13472390301525594,
+ "epoch": 7.360023378141438,
+ "grad_norm": 0.4053943157196045,
+ "learning_rate": 5.221428597784018e-05,
+ "loss": 0.07887161254882813,
+ "mean_token_accuracy": 0.9736681738495827,
+ "num_tokens": 9757913.0,
+ "step": 3150
+ },
+ {
+ "entropy": 0.14234945647418498,
+ "epoch": 7.4769140853302165,
+ "grad_norm": 0.46239086985588074,
+ "learning_rate": 4.7997081944331205e-05,
+ "loss": 0.08276874542236329,
+ "mean_token_accuracy": 0.9716356429457664,
+ "num_tokens": 9905790.0,
+ "step": 3200
+ },
+ {
+ "entropy": 0.13719632107764482,
+ "epoch": 7.593804792518995,
+ "grad_norm": 0.49838459491729736,
+ "learning_rate": 4.3919248174917856e-05,
+ "loss": 0.08003978729248047,
+ "mean_token_accuracy": 0.9731476292014122,
+ "num_tokens": 10064234.0,
+ "step": 3250
+ },
+ {
+ "entropy": 0.13751005962491036,
+ "epoch": 7.710695499707773,
+ "grad_norm": 0.4959999620914459,
+ "learning_rate": 3.9987564776352335e-05,
+ "loss": 0.07964549541473388,
+ "mean_token_accuracy": 0.9722868010401726,
+ "num_tokens": 10221089.0,
+ "step": 3300
+ },
+ {
+ "entropy": 0.13423183601349592,
+ "epoch": 7.827586206896552,
+ "grad_norm": 0.4889160692691803,
+ "learning_rate": 3.620856885501646e-05,
+ "loss": 0.07869738101959228,
+ "mean_token_accuracy": 0.973022837638855,
+ "num_tokens": 10379655.0,
+ "step": 3350
+ },
+ {
+ "entropy": 0.13587324120104313,
+ "epoch": 7.94447691408533,
+ "grad_norm": 0.4851946234703064,
+ "learning_rate": 3.258854364784726e-05,
+ "loss": 0.08004307746887207,
+ "mean_token_accuracy": 0.9723974686861038,
+ "num_tokens": 10531301.0,
+ "step": 3400
+ },
+ {
+ "epoch": 8.0,
+ "eval_entropy": 0.4391631047640528,
+ "eval_loss": 1.1321114301681519,
+ "eval_mean_token_accuracy": 0.7924294758152652,
+ "eval_num_tokens": 10605144.0,
+ "eval_runtime": 37.126,
+ "eval_samples_per_second": 33.157,
+ "eval_steps_per_second": 4.148,
+ "step": 3424
+ }
+ ],
+ "logging_steps": 50,
+ "max_steps": 4280,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 10,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.6381092874231296e+17,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3852/README.md b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3852/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3852/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3852/adapter_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3852/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9e45544de727016be134f25e606b7b2b87c67549
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3852/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.050044417519510255,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "v_proj",
+ "o_proj",
+ "q_proj",
+ "gate_proj",
+ "k_proj",
+ "down_proj",
+ "up_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3852/chat_template.jinja b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3852/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3852/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3852/tokenizer_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3852/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3852/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3852/trainer_state.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3852/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..aeac0e1485a5dfce2a7f1c470c2ee7b77e4dcdac
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-3852/trainer_state.json
@@ -0,0 +1,903 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 9.0,
+ "eval_steps": 500,
+ "global_step": 3852,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 1.8448095142841339,
+ "epoch": 0.11689070718877849,
+ "grad_norm": 1.4738068580627441,
+ "learning_rate": 3.0183098859435753e-05,
+ "loss": 1.6948321533203126,
+ "mean_token_accuracy": 0.6315939916670322,
+ "num_tokens": 153689.0,
+ "step": 50
+ },
+ {
+ "entropy": 0.9560942512750625,
+ "epoch": 0.23378141437755698,
+ "grad_norm": 1.378090739250183,
+ "learning_rate": 6.098217932824775e-05,
+ "loss": 0.8888477325439453,
+ "mean_token_accuracy": 0.7586692428588867,
+ "num_tokens": 310548.0,
+ "step": 100
+ },
+ {
+ "entropy": 0.848925698697567,
+ "epoch": 0.3506721215663355,
+ "grad_norm": 0.959621250629425,
+ "learning_rate": 9.178125979705974e-05,
+ "loss": 0.7850731658935547,
+ "mean_token_accuracy": 0.7774798944592476,
+ "num_tokens": 468560.0,
+ "step": 150
+ },
+ {
+ "entropy": 0.811761694252491,
+ "epoch": 0.46756282875511396,
+ "grad_norm": 0.8867514133453369,
+ "learning_rate": 0.00012258034026587175,
+ "loss": 0.7446884155273438,
+ "mean_token_accuracy": 0.7892421445250511,
+ "num_tokens": 619552.0,
+ "step": 200
+ },
+ {
+ "entropy": 0.7709847274422645,
+ "epoch": 0.5844535359438925,
+ "grad_norm": 0.5822142362594604,
+ "learning_rate": 0.00015337942073468373,
+ "loss": 0.7130480194091797,
+ "mean_token_accuracy": 0.7964023387432099,
+ "num_tokens": 771972.0,
+ "step": 250
+ },
+ {
+ "entropy": 0.7401632118225098,
+ "epoch": 0.701344243132671,
+ "grad_norm": 0.7557511925697327,
+ "learning_rate": 0.00018417850120349574,
+ "loss": 0.6857429504394531,
+ "mean_token_accuracy": 0.8041065946221352,
+ "num_tokens": 928088.0,
+ "step": 300
+ },
+ {
+ "entropy": 0.7295274111628532,
+ "epoch": 0.8182349503214494,
+ "grad_norm": 0.6769705414772034,
+ "learning_rate": 0.00021497758167230772,
+ "loss": 0.6801216125488281,
+ "mean_token_accuracy": 0.8037989619374275,
+ "num_tokens": 1086836.0,
+ "step": 350
+ },
+ {
+ "entropy": 0.710282272696495,
+ "epoch": 0.9351256575102279,
+ "grad_norm": 0.832009494304657,
+ "learning_rate": 0.00024577666214111967,
+ "loss": 0.6587177276611328,
+ "mean_token_accuracy": 0.8084024578332901,
+ "num_tokens": 1241227.0,
+ "step": 400
+ },
+ {
+ "epoch": 1.0,
+ "eval_entropy": 0.848237401672772,
+ "eval_loss": 0.7832363247871399,
+ "eval_mean_token_accuracy": 0.7817642971292719,
+ "eval_num_tokens": 1325643.0,
+ "eval_runtime": 36.8595,
+ "eval_samples_per_second": 33.397,
+ "eval_steps_per_second": 4.178,
+ "step": 428
+ },
+ {
+ "entropy": 0.7039205249230466,
+ "epoch": 1.0514319111630626,
+ "grad_norm": 0.6213153600692749,
+ "learning_rate": 0.00026362079548206905,
+ "loss": 0.6527813720703125,
+ "mean_token_accuracy": 0.8083000521564004,
+ "num_tokens": 1393922.0,
+ "step": 450
+ },
+ {
+ "entropy": 0.6749556151032448,
+ "epoch": 1.1683226183518411,
+ "grad_norm": 0.6444424390792847,
+ "learning_rate": 0.0002634191889632654,
+ "loss": 0.627038345336914,
+ "mean_token_accuracy": 0.8141268998384475,
+ "num_tokens": 1551442.0,
+ "step": 500
+ },
+ {
+ "entropy": 0.6727319830656051,
+ "epoch": 1.2852133255406195,
+ "grad_norm": 0.6360110640525818,
+ "learning_rate": 0.00026299877604786883,
+ "loss": 0.6232248687744141,
+ "mean_token_accuracy": 0.8153417313098907,
+ "num_tokens": 1706697.0,
+ "step": 550
+ },
+ {
+ "entropy": 0.6600116823613643,
+ "epoch": 1.4021040327293979,
+ "grad_norm": 0.46475180983543396,
+ "learning_rate": 0.00026236025574535523,
+ "loss": 0.6124750518798828,
+ "mean_token_accuracy": 0.8175293403863907,
+ "num_tokens": 1865246.0,
+ "step": 600
+ },
+ {
+ "entropy": 0.6494863349199295,
+ "epoch": 1.5189947399181765,
+ "grad_norm": 0.6132755875587463,
+ "learning_rate": 0.00026150468970660587,
+ "loss": 0.5997534942626953,
+ "mean_token_accuracy": 0.822050162255764,
+ "num_tokens": 2022060.0,
+ "step": 650
+ },
+ {
+ "entropy": 0.6511707258224487,
+ "epoch": 1.635885447106955,
+ "grad_norm": 0.570785641670227,
+ "learning_rate": 0.00026043350045872797,
+ "loss": 0.6059062576293945,
+ "mean_token_accuracy": 0.820236160159111,
+ "num_tokens": 2178909.0,
+ "step": 700
+ },
+ {
+ "entropy": 0.6550412529706955,
+ "epoch": 1.7527761542957334,
+ "grad_norm": 0.5398300290107727,
+ "learning_rate": 0.0002591484690398568,
+ "loss": 0.6031087875366211,
+ "mean_token_accuracy": 0.8206016793847084,
+ "num_tokens": 2329314.0,
+ "step": 750
+ },
+ {
+ "entropy": 0.6509851542115211,
+ "epoch": 1.869666861484512,
+ "grad_norm": 0.6913411617279053,
+ "learning_rate": 0.0002576517320378704,
+ "loss": 0.6023480224609375,
+ "mean_token_accuracy": 0.822216115295887,
+ "num_tokens": 2478559.0,
+ "step": 800
+ },
+ {
+ "entropy": 0.6560930109024048,
+ "epoch": 1.9865575686732906,
+ "grad_norm": 0.5568020939826965,
+ "learning_rate": 0.0002559457780379408,
+ "loss": 0.6031641006469727,
+ "mean_token_accuracy": 0.8205807811021805,
+ "num_tokens": 2631937.0,
+ "step": 850
+ },
+ {
+ "epoch": 2.0,
+ "eval_entropy": 0.794637631866839,
+ "eval_loss": 0.7390629649162292,
+ "eval_mean_token_accuracy": 0.7983863295672776,
+ "eval_num_tokens": 2651286.0,
+ "eval_runtime": 36.7346,
+ "eval_samples_per_second": 33.511,
+ "eval_steps_per_second": 4.192,
+ "step": 856
+ },
+ {
+ "entropy": 0.5709320336730037,
+ "epoch": 2.102863822326125,
+ "grad_norm": 0.4339260458946228,
+ "learning_rate": 0.0002540334434848298,
+ "loss": 0.5152394485473633,
+ "mean_token_accuracy": 0.8391561669919958,
+ "num_tokens": 2795200.0,
+ "step": 900
+ },
+ {
+ "entropy": 0.5768725140392781,
+ "epoch": 2.2197545295149035,
+ "grad_norm": 0.6345565319061279,
+ "learning_rate": 0.00025191790796680617,
+ "loss": 0.5241253662109375,
+ "mean_token_accuracy": 0.8377170553803444,
+ "num_tokens": 2950534.0,
+ "step": 950
+ },
+ {
+ "entropy": 0.570428695231676,
+ "epoch": 2.3366452367036823,
+ "grad_norm": 0.5838360786437988,
+ "learning_rate": 0.00024960268892902876,
+ "loss": 0.5187693405151367,
+ "mean_token_accuracy": 0.8382867723703384,
+ "num_tokens": 3101471.0,
+ "step": 1000
+ },
+ {
+ "entropy": 0.5713468404114246,
+ "epoch": 2.4535359438924607,
+ "grad_norm": 0.5641153454780579,
+ "learning_rate": 0.0002470916358251826,
+ "loss": 0.5234761810302735,
+ "mean_token_accuracy": 0.8382993021607399,
+ "num_tokens": 3258554.0,
+ "step": 1050
+ },
+ {
+ "entropy": 0.5645828676223755,
+ "epoch": 2.570426651081239,
+ "grad_norm": 0.504467785358429,
+ "learning_rate": 0.0002443889237170939,
+ "loss": 0.5156657409667968,
+ "mean_token_accuracy": 0.8391393664479255,
+ "num_tokens": 3416088.0,
+ "step": 1100
+ },
+ {
+ "entropy": 0.5659542319178581,
+ "epoch": 2.6873173582700174,
+ "grad_norm": 0.5225860476493835,
+ "learning_rate": 0.00024149904633296492,
+ "loss": 0.5180046844482422,
+ "mean_token_accuracy": 0.8389606699347496,
+ "num_tokens": 3567251.0,
+ "step": 1150
+ },
+ {
+ "entropy": 0.5713718670606613,
+ "epoch": 2.8042080654587958,
+ "grad_norm": 0.4986768066883087,
+ "learning_rate": 0.0002384268085957702,
+ "loss": 0.524637565612793,
+ "mean_token_accuracy": 0.8381215763092041,
+ "num_tokens": 3719464.0,
+ "step": 1200
+ },
+ {
+ "entropy": 0.5710382495820522,
+ "epoch": 2.9210987726475746,
+ "grad_norm": 0.5548401474952698,
+ "learning_rate": 0.00023517731863423788,
+ "loss": 0.5251939773559571,
+ "mean_token_accuracy": 0.8378310143947602,
+ "num_tokens": 3874423.0,
+ "step": 1250
+ },
+ {
+ "epoch": 3.0,
+ "eval_entropy": 0.6970683218209774,
+ "eval_loss": 0.7369498014450073,
+ "eval_mean_token_accuracy": 0.7919643076209278,
+ "eval_num_tokens": 3976929.0,
+ "eval_runtime": 36.9722,
+ "eval_samples_per_second": 33.295,
+ "eval_steps_per_second": 4.165,
+ "step": 1284
+ },
+ {
+ "entropy": 0.5407733828877684,
+ "epoch": 3.037405026300409,
+ "grad_norm": 0.6430732607841492,
+ "learning_rate": 0.00023175597928969878,
+ "loss": 0.49219757080078125,
+ "mean_token_accuracy": 0.8467060854686564,
+ "num_tokens": 4027609.0,
+ "step": 1300
+ },
+ {
+ "entropy": 0.4786584210395813,
+ "epoch": 3.1542957334891875,
+ "grad_norm": 0.6474543809890747,
+ "learning_rate": 0.00022816847913292434,
+ "loss": 0.4224985122680664,
+ "mean_token_accuracy": 0.8628773537278175,
+ "num_tokens": 4180560.0,
+ "step": 1350
+ },
+ {
+ "entropy": 0.49172899857163427,
+ "epoch": 3.2711864406779663,
+ "grad_norm": 0.549326479434967,
+ "learning_rate": 0.00022442078300588989,
+ "loss": 0.43718509674072265,
+ "mean_token_accuracy": 0.8583108323812485,
+ "num_tokens": 4329954.0,
+ "step": 1400
+ },
+ {
+ "entropy": 0.4829080794751644,
+ "epoch": 3.3880771478667446,
+ "grad_norm": 0.5274888873100281,
+ "learning_rate": 0.00022051912210418922,
+ "loss": 0.4315507125854492,
+ "mean_token_accuracy": 0.8612637943029404,
+ "num_tokens": 4484354.0,
+ "step": 1450
+ },
+ {
+ "entropy": 0.4855460473895073,
+ "epoch": 3.504967855055523,
+ "grad_norm": 0.5846306681632996,
+ "learning_rate": 0.00021646998361658985,
+ "loss": 0.43724521636962893,
+ "mean_token_accuracy": 0.8589270997047425,
+ "num_tokens": 4638919.0,
+ "step": 1500
+ },
+ {
+ "entropy": 0.4905838692188263,
+ "epoch": 3.6218585622443014,
+ "grad_norm": 0.5149123072624207,
+ "learning_rate": 0.00021228009993895495,
+ "loss": 0.44197372436523436,
+ "mean_token_accuracy": 0.8584243017435074,
+ "num_tokens": 4795540.0,
+ "step": 1550
+ },
+ {
+ "entropy": 0.4994237454235554,
+ "epoch": 3.73874926943308,
+ "grad_norm": 0.5417225956916809,
+ "learning_rate": 0.0002079564374804665,
+ "loss": 0.4491108322143555,
+ "mean_token_accuracy": 0.8556393891572952,
+ "num_tokens": 4951177.0,
+ "step": 1600
+ },
+ {
+ "entropy": 0.49903531357645986,
+ "epoch": 3.8556399766218585,
+ "grad_norm": 0.5086734890937805,
+ "learning_rate": 0.00020350618508076015,
+ "loss": 0.4518918228149414,
+ "mean_token_accuracy": 0.8549707370996475,
+ "num_tokens": 5104814.0,
+ "step": 1650
+ },
+ {
+ "entropy": 0.4859523306787014,
+ "epoch": 3.972530683810637,
+ "grad_norm": 0.5446717739105225,
+ "learning_rate": 0.00019893674205723094,
+ "loss": 0.4410313415527344,
+ "mean_token_accuracy": 0.8587654489278793,
+ "num_tokens": 5265185.0,
+ "step": 1700
+ },
+ {
+ "epoch": 4.0,
+ "eval_entropy": 0.6577693150415049,
+ "eval_loss": 0.7380754947662354,
+ "eval_mean_token_accuracy": 0.8039549565934515,
+ "eval_num_tokens": 5302572.0,
+ "eval_runtime": 36.5757,
+ "eval_samples_per_second": 33.656,
+ "eval_steps_per_second": 4.21,
+ "step": 1712
+ },
+ {
+ "entropy": 0.41945636961328325,
+ "epoch": 4.088836937463472,
+ "grad_norm": 0.5925881266593933,
+ "learning_rate": 0.00019425570590238332,
+ "loss": 0.3575031280517578,
+ "mean_token_accuracy": 0.8812823918596584,
+ "num_tokens": 5419002.0,
+ "step": 1750
+ },
+ {
+ "entropy": 0.40106018364429474,
+ "epoch": 4.20572764465225,
+ "grad_norm": 0.7094829082489014,
+ "learning_rate": 0.0001894708596516806,
+ "loss": 0.3372145080566406,
+ "mean_token_accuracy": 0.8856011089682579,
+ "num_tokens": 5572969.0,
+ "step": 1800
+ },
+ {
+ "entropy": 0.4121378688514233,
+ "epoch": 4.322618351841029,
+ "grad_norm": 0.6077721118927002,
+ "learning_rate": 0.00018459015894289695,
+ "loss": 0.34956165313720705,
+ "mean_token_accuracy": 0.8818695649504662,
+ "num_tokens": 5733805.0,
+ "step": 1850
+ },
+ {
+ "entropy": 0.40683484479784965,
+ "epoch": 4.439509059029807,
+ "grad_norm": 0.5614883303642273,
+ "learning_rate": 0.00017962171878848807,
+ "loss": 0.34749691009521483,
+ "mean_token_accuracy": 0.882930768430233,
+ "num_tokens": 5888694.0,
+ "step": 1900
+ },
+ {
+ "entropy": 0.4100585526227951,
+ "epoch": 4.556399766218585,
+ "grad_norm": 0.5506443977355957,
+ "learning_rate": 0.00017457380008297383,
+ "loss": 0.3501173782348633,
+ "mean_token_accuracy": 0.8827524068951607,
+ "num_tokens": 6042655.0,
+ "step": 1950
+ },
+ {
+ "entropy": 0.41173958137631417,
+ "epoch": 4.673290473407365,
+ "grad_norm": 0.5409566164016724,
+ "learning_rate": 0.00016945479586776672,
+ "loss": 0.3541391754150391,
+ "mean_token_accuracy": 0.8808356913924217,
+ "num_tokens": 6203980.0,
+ "step": 2000
+ },
+ {
+ "entropy": 0.4158783262968063,
+ "epoch": 4.790181180596143,
+ "grad_norm": 0.572618842124939,
+ "learning_rate": 0.00016427321737628295,
+ "loss": 0.3536404800415039,
+ "mean_token_accuracy": 0.880456589460373,
+ "num_tokens": 6354604.0,
+ "step": 2050
+ },
+ {
+ "entropy": 0.408166281580925,
+ "epoch": 4.907071887784921,
+ "grad_norm": 0.5987762808799744,
+ "learning_rate": 0.00015903767988253897,
+ "loss": 0.3518599319458008,
+ "mean_token_accuracy": 0.8815453574061394,
+ "num_tokens": 6507484.0,
+ "step": 2100
+ },
+ {
+ "epoch": 5.0,
+ "eval_entropy": 0.6047254425364655,
+ "eval_loss": 0.7816920876502991,
+ "eval_mean_token_accuracy": 0.7987685733801358,
+ "eval_num_tokens": 6628215.0,
+ "eval_runtime": 36.9374,
+ "eval_samples_per_second": 33.327,
+ "eval_steps_per_second": 4.169,
+ "step": 2140
+ },
+ {
+ "entropy": 0.38985791001188097,
+ "epoch": 5.023378141437756,
+ "grad_norm": 0.7454882860183716,
+ "learning_rate": 0.00015375688837676246,
+ "loss": 0.32916854858398437,
+ "mean_token_accuracy": 0.8886270373310875,
+ "num_tokens": 6662015.0,
+ "step": 2150
+ },
+ {
+ "entropy": 0.30170531339943407,
+ "epoch": 5.140268848626534,
+ "grad_norm": 0.5329811573028564,
+ "learning_rate": 0.0001484396230918344,
+ "loss": 0.23112209320068358,
+ "mean_token_accuracy": 0.9198204749822616,
+ "num_tokens": 6817989.0,
+ "step": 2200
+ },
+ {
+ "entropy": 0.3075008810311556,
+ "epoch": 5.257159555815313,
+ "grad_norm": 0.6896093487739563,
+ "learning_rate": 0.00014309472490462715,
+ "loss": 0.2416652488708496,
+ "mean_token_accuracy": 0.9158366405963898,
+ "num_tokens": 6977021.0,
+ "step": 2250
+ },
+ {
+ "entropy": 0.31153449676930905,
+ "epoch": 5.374050263004091,
+ "grad_norm": 0.6930833458900452,
+ "learning_rate": 0.00013773108063651161,
+ "loss": 0.24753889083862304,
+ "mean_token_accuracy": 0.9136614826321602,
+ "num_tokens": 7133360.0,
+ "step": 2300
+ },
+ {
+ "entropy": 0.3154346436262131,
+ "epoch": 5.490940970192869,
+ "grad_norm": 0.6476582884788513,
+ "learning_rate": 0.0001323576082774734,
+ "loss": 0.24872392654418946,
+ "mean_token_accuracy": 0.9135914054512978,
+ "num_tokens": 7282372.0,
+ "step": 2350
+ },
+ {
+ "entropy": 0.313641587048769,
+ "epoch": 5.607831677381649,
+ "grad_norm": 0.6953482627868652,
+ "learning_rate": 0.0001269832421584059,
+ "loss": 0.24919464111328124,
+ "mean_token_accuracy": 0.9131662029027939,
+ "num_tokens": 7435766.0,
+ "step": 2400
+ },
+ {
+ "entropy": 0.3163894522190094,
+ "epoch": 5.724722384570427,
+ "grad_norm": 0.6673940420150757,
+ "learning_rate": 0.00012161691809623384,
+ "loss": 0.25111726760864256,
+ "mean_token_accuracy": 0.913602537214756,
+ "num_tokens": 7578408.0,
+ "step": 2450
+ },
+ {
+ "entropy": 0.3183368914574385,
+ "epoch": 5.841613091759205,
+ "grad_norm": 0.7174487709999084,
+ "learning_rate": 0.00011626755853656576,
+ "loss": 0.2518378448486328,
+ "mean_token_accuracy": 0.912000589966774,
+ "num_tokens": 7734599.0,
+ "step": 2500
+ },
+ {
+ "entropy": 0.3077564483135939,
+ "epoch": 5.958503798947984,
+ "grad_norm": 0.6662200689315796,
+ "learning_rate": 0.00011094405771857891,
+ "loss": 0.24354087829589843,
+ "mean_token_accuracy": 0.9140791621804237,
+ "num_tokens": 7898028.0,
+ "step": 2550
+ },
+ {
+ "epoch": 6.0,
+ "eval_entropy": 0.5403795917699863,
+ "eval_loss": 0.8572548627853394,
+ "eval_mean_token_accuracy": 0.796143418395674,
+ "eval_num_tokens": 7953858.0,
+ "eval_runtime": 36.7176,
+ "eval_samples_per_second": 33.526,
+ "eval_steps_per_second": 4.194,
+ "step": 2568
+ },
+ {
+ "entropy": 0.25110363144071857,
+ "epoch": 6.074810052600818,
+ "grad_norm": 0.6466763019561768,
+ "learning_rate": 0.0001056552668868023,
+ "loss": 0.1814822769165039,
+ "mean_token_accuracy": 0.9370300626035911,
+ "num_tokens": 8052646.0,
+ "step": 2600
+ },
+ {
+ "entropy": 0.2063255251944065,
+ "epoch": 6.1917007597895966,
+ "grad_norm": 0.6950973272323608,
+ "learning_rate": 0.0001004099795743853,
+ "loss": 0.13960565567016603,
+ "mean_token_accuracy": 0.9511301657557487,
+ "num_tokens": 8212323.0,
+ "step": 2650
+ },
+ {
+ "entropy": 0.20775293290615082,
+ "epoch": 6.308591466978375,
+ "grad_norm": 0.660478413105011,
+ "learning_rate": 9.52169169823223e-05,
+ "loss": 0.14319146156311036,
+ "mean_token_accuracy": 0.9500676044821739,
+ "num_tokens": 8366715.0,
+ "step": 2700
+ },
+ {
+ "entropy": 0.20977566957473756,
+ "epoch": 6.425482174167154,
+ "grad_norm": 0.6141477823257446,
+ "learning_rate": 9.00847134789411e-05,
+ "loss": 0.14525352478027342,
+ "mean_token_accuracy": 0.9490846493840217,
+ "num_tokens": 8515955.0,
+ "step": 2750
+ },
+ {
+ "entropy": 0.2143413795530796,
+ "epoch": 6.5423728813559325,
+ "grad_norm": 0.612184464931488,
+ "learning_rate": 8.502190224376661e-05,
+ "loss": 0.14751774787902833,
+ "mean_token_accuracy": 0.9478235357999801,
+ "num_tokens": 8672361.0,
+ "step": 2800
+ },
+ {
+ "entropy": 0.21027401253581046,
+ "epoch": 6.659263588544711,
+ "grad_norm": 0.644019365310669,
+ "learning_rate": 8.003690107962787e-05,
+ "loss": 0.1495517921447754,
+ "mean_token_accuracy": 0.9470956793427467,
+ "num_tokens": 8827819.0,
+ "step": 2850
+ },
+ {
+ "entropy": 0.2109992092102766,
+ "epoch": 6.776154295733489,
+ "grad_norm": 0.5456500053405762,
+ "learning_rate": 7.513799841659927e-05,
+ "loss": 0.14893126487731934,
+ "mean_token_accuracy": 0.9479587042331695,
+ "num_tokens": 8979535.0,
+ "step": 2900
+ },
+ {
+ "entropy": 0.20949276342988013,
+ "epoch": 6.893045002922268,
+ "grad_norm": 0.6803008913993835,
+ "learning_rate": 7.033333953104615e-05,
+ "loss": 0.1461741542816162,
+ "mean_token_accuracy": 0.9480843073129654,
+ "num_tokens": 9134134.0,
+ "step": 2950
+ },
+ {
+ "epoch": 7.0,
+ "eval_entropy": 0.4759114793949313,
+ "eval_loss": 0.9827883839607239,
+ "eval_mean_token_accuracy": 0.7945680889216337,
+ "eval_num_tokens": 9279501.0,
+ "eval_runtime": 36.4188,
+ "eval_samples_per_second": 33.801,
+ "eval_steps_per_second": 4.229,
+ "step": 2996
+ },
+ {
+ "entropy": 0.20382985060838,
+ "epoch": 7.009351256575102,
+ "grad_norm": 0.42079976201057434,
+ "learning_rate": 6.563091300268857e-05,
+ "loss": 0.13976585388183593,
+ "mean_token_accuracy": 0.9511888485458029,
+ "num_tokens": 9290416.0,
+ "step": 3000
+ },
+ {
+ "entropy": 0.14199006125330926,
+ "epoch": 7.1262419637638805,
+ "grad_norm": 0.6262232065200806,
+ "learning_rate": 6.103853743220019e-05,
+ "loss": 0.08045654296875,
+ "mean_token_accuracy": 0.9727814987301826,
+ "num_tokens": 9444705.0,
+ "step": 3050
+ },
+ {
+ "entropy": 0.13681559398770332,
+ "epoch": 7.243132670952659,
+ "grad_norm": 0.6460694074630737,
+ "learning_rate": 5.6563848441427006e-05,
+ "loss": 0.0795314359664917,
+ "mean_token_accuracy": 0.973130469918251,
+ "num_tokens": 9601476.0,
+ "step": 3100
+ },
+ {
+ "entropy": 0.13472390301525594,
+ "epoch": 7.360023378141438,
+ "grad_norm": 0.4053943157196045,
+ "learning_rate": 5.221428597784018e-05,
+ "loss": 0.07887161254882813,
+ "mean_token_accuracy": 0.9736681738495827,
+ "num_tokens": 9757913.0,
+ "step": 3150
+ },
+ {
+ "entropy": 0.14234945647418498,
+ "epoch": 7.4769140853302165,
+ "grad_norm": 0.46239086985588074,
+ "learning_rate": 4.7997081944331205e-05,
+ "loss": 0.08276874542236329,
+ "mean_token_accuracy": 0.9716356429457664,
+ "num_tokens": 9905790.0,
+ "step": 3200
+ },
+ {
+ "entropy": 0.13719632107764482,
+ "epoch": 7.593804792518995,
+ "grad_norm": 0.49838459491729736,
+ "learning_rate": 4.3919248174917856e-05,
+ "loss": 0.08003978729248047,
+ "mean_token_accuracy": 0.9731476292014122,
+ "num_tokens": 10064234.0,
+ "step": 3250
+ },
+ {
+ "entropy": 0.13751005962491036,
+ "epoch": 7.710695499707773,
+ "grad_norm": 0.4959999620914459,
+ "learning_rate": 3.9987564776352335e-05,
+ "loss": 0.07964549541473388,
+ "mean_token_accuracy": 0.9722868010401726,
+ "num_tokens": 10221089.0,
+ "step": 3300
+ },
+ {
+ "entropy": 0.13423183601349592,
+ "epoch": 7.827586206896552,
+ "grad_norm": 0.4889160692691803,
+ "learning_rate": 3.620856885501646e-05,
+ "loss": 0.07869738101959228,
+ "mean_token_accuracy": 0.973022837638855,
+ "num_tokens": 10379655.0,
+ "step": 3350
+ },
+ {
+ "entropy": 0.13587324120104313,
+ "epoch": 7.94447691408533,
+ "grad_norm": 0.4851946234703064,
+ "learning_rate": 3.258854364784726e-05,
+ "loss": 0.08004307746887207,
+ "mean_token_accuracy": 0.9723974686861038,
+ "num_tokens": 10531301.0,
+ "step": 3400
+ },
+ {
+ "epoch": 8.0,
+ "eval_entropy": 0.4391631047640528,
+ "eval_loss": 1.1321114301681519,
+ "eval_mean_token_accuracy": 0.7924294758152652,
+ "eval_num_tokens": 10605144.0,
+ "eval_runtime": 37.126,
+ "eval_samples_per_second": 33.157,
+ "eval_steps_per_second": 4.148,
+ "step": 3424
+ },
+ {
+ "entropy": 0.12316639324528487,
+ "epoch": 8.060783167738165,
+ "grad_norm": 0.3510015606880188,
+ "learning_rate": 2.9133508075364238e-05,
+ "loss": 0.06677456378936768,
+ "mean_token_accuracy": 0.9770565955483135,
+ "num_tokens": 10686079.0,
+ "step": 3450
+ },
+ {
+ "entropy": 0.11033655313774943,
+ "epoch": 8.177673874926944,
+ "grad_norm": 0.27871957421302795,
+ "learning_rate": 2.5849206734169147e-05,
+ "loss": 0.05629729747772217,
+ "mean_token_accuracy": 0.9804861381649971,
+ "num_tokens": 10838236.0,
+ "step": 3500
+ },
+ {
+ "entropy": 0.10826377695426345,
+ "epoch": 8.294564582115722,
+ "grad_norm": 0.44871190190315247,
+ "learning_rate": 2.2741100345556444e-05,
+ "loss": 0.05603146076202393,
+ "mean_token_accuracy": 0.9803140532970428,
+ "num_tokens": 10989653.0,
+ "step": 3550
+ },
+ {
+ "entropy": 0.10614268690347671,
+ "epoch": 8.4114552893045,
+ "grad_norm": 0.25647690892219543,
+ "learning_rate": 1.9814356676115743e-05,
+ "loss": 0.0554099178314209,
+ "mean_token_accuracy": 0.9802037307620048,
+ "num_tokens": 11144096.0,
+ "step": 3600
+ },
+ {
+ "entropy": 0.10369085857644678,
+ "epoch": 8.528345996493279,
+ "grad_norm": 0.32260316610336304,
+ "learning_rate": 1.7073841945422747e-05,
+ "loss": 0.05409110546112061,
+ "mean_token_accuracy": 0.980969468653202,
+ "num_tokens": 11303465.0,
+ "step": 3650
+ },
+ {
+ "entropy": 0.10663174549117685,
+ "epoch": 8.645236703682057,
+ "grad_norm": 0.33514541387557983,
+ "learning_rate": 1.4524112735104158e-05,
+ "loss": 0.056715693473815915,
+ "mean_token_accuracy": 0.9798875021934509,
+ "num_tokens": 11455533.0,
+ "step": 3700
+ },
+ {
+ "entropy": 0.1048200499266386,
+ "epoch": 8.762127410870836,
+ "grad_norm": 0.3340476453304291,
+ "learning_rate": 1.2169408412729229e-05,
+ "loss": 0.05345354557037354,
+ "mean_token_accuracy": 0.9803790295124054,
+ "num_tokens": 11618619.0,
+ "step": 3750
+ },
+ {
+ "entropy": 0.10449992345646024,
+ "epoch": 8.879018118059614,
+ "grad_norm": 0.30406567454338074,
+ "learning_rate": 1.0013644083125266e-05,
+ "loss": 0.054877762794494626,
+ "mean_token_accuracy": 0.9807043489813805,
+ "num_tokens": 11774627.0,
+ "step": 3800
+ },
+ {
+ "entropy": 0.10615571580827236,
+ "epoch": 8.995908825248392,
+ "grad_norm": 0.36648011207580566,
+ "learning_rate": 8.060404078835946e-06,
+ "loss": 0.056678996086120606,
+ "mean_token_accuracy": 0.9797374567389489,
+ "num_tokens": 11925408.0,
+ "step": 3850
+ },
+ {
+ "epoch": 9.0,
+ "eval_entropy": 0.4037178186820699,
+ "eval_loss": 1.254728078842163,
+ "eval_mean_token_accuracy": 0.7945115701718763,
+ "eval_num_tokens": 11930787.0,
+ "eval_runtime": 36.9056,
+ "eval_samples_per_second": 33.355,
+ "eval_steps_per_second": 4.173,
+ "step": 3852
+ }
+ ],
+ "logging_steps": 50,
+ "max_steps": 4280,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 10,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.84246098288768e+17,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-428/README.md b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-428/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-428/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-428/adapter_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-428/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9e45544de727016be134f25e606b7b2b87c67549
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-428/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.050044417519510255,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "v_proj",
+ "o_proj",
+ "q_proj",
+ "gate_proj",
+ "k_proj",
+ "down_proj",
+ "up_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-428/chat_template.jinja b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-428/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-428/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-428/tokenizer_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-428/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-428/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-428/trainer_state.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-428/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c6fa236661f43829aae12f37410d042a38dd8dcd
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-428/trainer_state.json
@@ -0,0 +1,125 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 1.0,
+ "eval_steps": 500,
+ "global_step": 428,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 1.8448095142841339,
+ "epoch": 0.11689070718877849,
+ "grad_norm": 1.4738068580627441,
+ "learning_rate": 3.0183098859435753e-05,
+ "loss": 1.6948321533203126,
+ "mean_token_accuracy": 0.6315939916670322,
+ "num_tokens": 153689.0,
+ "step": 50
+ },
+ {
+ "entropy": 0.9560942512750625,
+ "epoch": 0.23378141437755698,
+ "grad_norm": 1.378090739250183,
+ "learning_rate": 6.098217932824775e-05,
+ "loss": 0.8888477325439453,
+ "mean_token_accuracy": 0.7586692428588867,
+ "num_tokens": 310548.0,
+ "step": 100
+ },
+ {
+ "entropy": 0.848925698697567,
+ "epoch": 0.3506721215663355,
+ "grad_norm": 0.959621250629425,
+ "learning_rate": 9.178125979705974e-05,
+ "loss": 0.7850731658935547,
+ "mean_token_accuracy": 0.7774798944592476,
+ "num_tokens": 468560.0,
+ "step": 150
+ },
+ {
+ "entropy": 0.811761694252491,
+ "epoch": 0.46756282875511396,
+ "grad_norm": 0.8867514133453369,
+ "learning_rate": 0.00012258034026587175,
+ "loss": 0.7446884155273438,
+ "mean_token_accuracy": 0.7892421445250511,
+ "num_tokens": 619552.0,
+ "step": 200
+ },
+ {
+ "entropy": 0.7709847274422645,
+ "epoch": 0.5844535359438925,
+ "grad_norm": 0.5822142362594604,
+ "learning_rate": 0.00015337942073468373,
+ "loss": 0.7130480194091797,
+ "mean_token_accuracy": 0.7964023387432099,
+ "num_tokens": 771972.0,
+ "step": 250
+ },
+ {
+ "entropy": 0.7401632118225098,
+ "epoch": 0.701344243132671,
+ "grad_norm": 0.7557511925697327,
+ "learning_rate": 0.00018417850120349574,
+ "loss": 0.6857429504394531,
+ "mean_token_accuracy": 0.8041065946221352,
+ "num_tokens": 928088.0,
+ "step": 300
+ },
+ {
+ "entropy": 0.7295274111628532,
+ "epoch": 0.8182349503214494,
+ "grad_norm": 0.6769705414772034,
+ "learning_rate": 0.00021497758167230772,
+ "loss": 0.6801216125488281,
+ "mean_token_accuracy": 0.8037989619374275,
+ "num_tokens": 1086836.0,
+ "step": 350
+ },
+ {
+ "entropy": 0.710282272696495,
+ "epoch": 0.9351256575102279,
+ "grad_norm": 0.832009494304657,
+ "learning_rate": 0.00024577666214111967,
+ "loss": 0.6587177276611328,
+ "mean_token_accuracy": 0.8084024578332901,
+ "num_tokens": 1241227.0,
+ "step": 400
+ },
+ {
+ "epoch": 1.0,
+ "eval_entropy": 0.848237401672772,
+ "eval_loss": 0.7832363247871399,
+ "eval_mean_token_accuracy": 0.7817642971292719,
+ "eval_num_tokens": 1325643.0,
+ "eval_runtime": 36.8595,
+ "eval_samples_per_second": 33.397,
+ "eval_steps_per_second": 4.178,
+ "step": 428
+ }
+ ],
+ "logging_steps": 50,
+ "max_steps": 4280,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 10,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 2.053230260741376e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-4280/README.md b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-4280/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-4280/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-4280/adapter_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-4280/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9e45544de727016be134f25e606b7b2b87c67549
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-4280/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.050044417519510255,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "v_proj",
+ "o_proj",
+ "q_proj",
+ "gate_proj",
+ "k_proj",
+ "down_proj",
+ "up_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-4280/chat_template.jinja b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-4280/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-4280/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-4280/tokenizer_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-4280/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-4280/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-4280/trainer_state.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-4280/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..307d039bfa0c3dffaa766861956197041f3bc819
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-4280/trainer_state.json
@@ -0,0 +1,994 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 10.0,
+ "eval_steps": 500,
+ "global_step": 4280,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 1.8448095142841339,
+ "epoch": 0.11689070718877849,
+ "grad_norm": 1.4738068580627441,
+ "learning_rate": 3.0183098859435753e-05,
+ "loss": 1.6948321533203126,
+ "mean_token_accuracy": 0.6315939916670322,
+ "num_tokens": 153689.0,
+ "step": 50
+ },
+ {
+ "entropy": 0.9560942512750625,
+ "epoch": 0.23378141437755698,
+ "grad_norm": 1.378090739250183,
+ "learning_rate": 6.098217932824775e-05,
+ "loss": 0.8888477325439453,
+ "mean_token_accuracy": 0.7586692428588867,
+ "num_tokens": 310548.0,
+ "step": 100
+ },
+ {
+ "entropy": 0.848925698697567,
+ "epoch": 0.3506721215663355,
+ "grad_norm": 0.959621250629425,
+ "learning_rate": 9.178125979705974e-05,
+ "loss": 0.7850731658935547,
+ "mean_token_accuracy": 0.7774798944592476,
+ "num_tokens": 468560.0,
+ "step": 150
+ },
+ {
+ "entropy": 0.811761694252491,
+ "epoch": 0.46756282875511396,
+ "grad_norm": 0.8867514133453369,
+ "learning_rate": 0.00012258034026587175,
+ "loss": 0.7446884155273438,
+ "mean_token_accuracy": 0.7892421445250511,
+ "num_tokens": 619552.0,
+ "step": 200
+ },
+ {
+ "entropy": 0.7709847274422645,
+ "epoch": 0.5844535359438925,
+ "grad_norm": 0.5822142362594604,
+ "learning_rate": 0.00015337942073468373,
+ "loss": 0.7130480194091797,
+ "mean_token_accuracy": 0.7964023387432099,
+ "num_tokens": 771972.0,
+ "step": 250
+ },
+ {
+ "entropy": 0.7401632118225098,
+ "epoch": 0.701344243132671,
+ "grad_norm": 0.7557511925697327,
+ "learning_rate": 0.00018417850120349574,
+ "loss": 0.6857429504394531,
+ "mean_token_accuracy": 0.8041065946221352,
+ "num_tokens": 928088.0,
+ "step": 300
+ },
+ {
+ "entropy": 0.7295274111628532,
+ "epoch": 0.8182349503214494,
+ "grad_norm": 0.6769705414772034,
+ "learning_rate": 0.00021497758167230772,
+ "loss": 0.6801216125488281,
+ "mean_token_accuracy": 0.8037989619374275,
+ "num_tokens": 1086836.0,
+ "step": 350
+ },
+ {
+ "entropy": 0.710282272696495,
+ "epoch": 0.9351256575102279,
+ "grad_norm": 0.832009494304657,
+ "learning_rate": 0.00024577666214111967,
+ "loss": 0.6587177276611328,
+ "mean_token_accuracy": 0.8084024578332901,
+ "num_tokens": 1241227.0,
+ "step": 400
+ },
+ {
+ "epoch": 1.0,
+ "eval_entropy": 0.848237401672772,
+ "eval_loss": 0.7832363247871399,
+ "eval_mean_token_accuracy": 0.7817642971292719,
+ "eval_num_tokens": 1325643.0,
+ "eval_runtime": 36.8595,
+ "eval_samples_per_second": 33.397,
+ "eval_steps_per_second": 4.178,
+ "step": 428
+ },
+ {
+ "entropy": 0.7039205249230466,
+ "epoch": 1.0514319111630626,
+ "grad_norm": 0.6213153600692749,
+ "learning_rate": 0.00026362079548206905,
+ "loss": 0.6527813720703125,
+ "mean_token_accuracy": 0.8083000521564004,
+ "num_tokens": 1393922.0,
+ "step": 450
+ },
+ {
+ "entropy": 0.6749556151032448,
+ "epoch": 1.1683226183518411,
+ "grad_norm": 0.6444424390792847,
+ "learning_rate": 0.0002634191889632654,
+ "loss": 0.627038345336914,
+ "mean_token_accuracy": 0.8141268998384475,
+ "num_tokens": 1551442.0,
+ "step": 500
+ },
+ {
+ "entropy": 0.6727319830656051,
+ "epoch": 1.2852133255406195,
+ "grad_norm": 0.6360110640525818,
+ "learning_rate": 0.00026299877604786883,
+ "loss": 0.6232248687744141,
+ "mean_token_accuracy": 0.8153417313098907,
+ "num_tokens": 1706697.0,
+ "step": 550
+ },
+ {
+ "entropy": 0.6600116823613643,
+ "epoch": 1.4021040327293979,
+ "grad_norm": 0.46475180983543396,
+ "learning_rate": 0.00026236025574535523,
+ "loss": 0.6124750518798828,
+ "mean_token_accuracy": 0.8175293403863907,
+ "num_tokens": 1865246.0,
+ "step": 600
+ },
+ {
+ "entropy": 0.6494863349199295,
+ "epoch": 1.5189947399181765,
+ "grad_norm": 0.6132755875587463,
+ "learning_rate": 0.00026150468970660587,
+ "loss": 0.5997534942626953,
+ "mean_token_accuracy": 0.822050162255764,
+ "num_tokens": 2022060.0,
+ "step": 650
+ },
+ {
+ "entropy": 0.6511707258224487,
+ "epoch": 1.635885447106955,
+ "grad_norm": 0.570785641670227,
+ "learning_rate": 0.00026043350045872797,
+ "loss": 0.6059062576293945,
+ "mean_token_accuracy": 0.820236160159111,
+ "num_tokens": 2178909.0,
+ "step": 700
+ },
+ {
+ "entropy": 0.6550412529706955,
+ "epoch": 1.7527761542957334,
+ "grad_norm": 0.5398300290107727,
+ "learning_rate": 0.0002591484690398568,
+ "loss": 0.6031087875366211,
+ "mean_token_accuracy": 0.8206016793847084,
+ "num_tokens": 2329314.0,
+ "step": 750
+ },
+ {
+ "entropy": 0.6509851542115211,
+ "epoch": 1.869666861484512,
+ "grad_norm": 0.6913411617279053,
+ "learning_rate": 0.0002576517320378704,
+ "loss": 0.6023480224609375,
+ "mean_token_accuracy": 0.822216115295887,
+ "num_tokens": 2478559.0,
+ "step": 800
+ },
+ {
+ "entropy": 0.6560930109024048,
+ "epoch": 1.9865575686732906,
+ "grad_norm": 0.5568020939826965,
+ "learning_rate": 0.0002559457780379408,
+ "loss": 0.6031641006469727,
+ "mean_token_accuracy": 0.8205807811021805,
+ "num_tokens": 2631937.0,
+ "step": 850
+ },
+ {
+ "epoch": 2.0,
+ "eval_entropy": 0.794637631866839,
+ "eval_loss": 0.7390629649162292,
+ "eval_mean_token_accuracy": 0.7983863295672776,
+ "eval_num_tokens": 2651286.0,
+ "eval_runtime": 36.7346,
+ "eval_samples_per_second": 33.511,
+ "eval_steps_per_second": 4.192,
+ "step": 856
+ },
+ {
+ "entropy": 0.5709320336730037,
+ "epoch": 2.102863822326125,
+ "grad_norm": 0.4339260458946228,
+ "learning_rate": 0.0002540334434848298,
+ "loss": 0.5152394485473633,
+ "mean_token_accuracy": 0.8391561669919958,
+ "num_tokens": 2795200.0,
+ "step": 900
+ },
+ {
+ "entropy": 0.5768725140392781,
+ "epoch": 2.2197545295149035,
+ "grad_norm": 0.6345565319061279,
+ "learning_rate": 0.00025191790796680617,
+ "loss": 0.5241253662109375,
+ "mean_token_accuracy": 0.8377170553803444,
+ "num_tokens": 2950534.0,
+ "step": 950
+ },
+ {
+ "entropy": 0.570428695231676,
+ "epoch": 2.3366452367036823,
+ "grad_norm": 0.5838360786437988,
+ "learning_rate": 0.00024960268892902876,
+ "loss": 0.5187693405151367,
+ "mean_token_accuracy": 0.8382867723703384,
+ "num_tokens": 3101471.0,
+ "step": 1000
+ },
+ {
+ "entropy": 0.5713468404114246,
+ "epoch": 2.4535359438924607,
+ "grad_norm": 0.5641153454780579,
+ "learning_rate": 0.0002470916358251826,
+ "loss": 0.5234761810302735,
+ "mean_token_accuracy": 0.8382993021607399,
+ "num_tokens": 3258554.0,
+ "step": 1050
+ },
+ {
+ "entropy": 0.5645828676223755,
+ "epoch": 2.570426651081239,
+ "grad_norm": 0.504467785358429,
+ "learning_rate": 0.0002443889237170939,
+ "loss": 0.5156657409667968,
+ "mean_token_accuracy": 0.8391393664479255,
+ "num_tokens": 3416088.0,
+ "step": 1100
+ },
+ {
+ "entropy": 0.5659542319178581,
+ "epoch": 2.6873173582700174,
+ "grad_norm": 0.5225860476493835,
+ "learning_rate": 0.00024149904633296492,
+ "loss": 0.5180046844482422,
+ "mean_token_accuracy": 0.8389606699347496,
+ "num_tokens": 3567251.0,
+ "step": 1150
+ },
+ {
+ "entropy": 0.5713718670606613,
+ "epoch": 2.8042080654587958,
+ "grad_norm": 0.4986768066883087,
+ "learning_rate": 0.0002384268085957702,
+ "loss": 0.524637565612793,
+ "mean_token_accuracy": 0.8381215763092041,
+ "num_tokens": 3719464.0,
+ "step": 1200
+ },
+ {
+ "entropy": 0.5710382495820522,
+ "epoch": 2.9210987726475746,
+ "grad_norm": 0.5548401474952698,
+ "learning_rate": 0.00023517731863423788,
+ "loss": 0.5251939773559571,
+ "mean_token_accuracy": 0.8378310143947602,
+ "num_tokens": 3874423.0,
+ "step": 1250
+ },
+ {
+ "epoch": 3.0,
+ "eval_entropy": 0.6970683218209774,
+ "eval_loss": 0.7369498014450073,
+ "eval_mean_token_accuracy": 0.7919643076209278,
+ "eval_num_tokens": 3976929.0,
+ "eval_runtime": 36.9722,
+ "eval_samples_per_second": 33.295,
+ "eval_steps_per_second": 4.165,
+ "step": 1284
+ },
+ {
+ "entropy": 0.5407733828877684,
+ "epoch": 3.037405026300409,
+ "grad_norm": 0.6430732607841492,
+ "learning_rate": 0.00023175597928969878,
+ "loss": 0.49219757080078125,
+ "mean_token_accuracy": 0.8467060854686564,
+ "num_tokens": 4027609.0,
+ "step": 1300
+ },
+ {
+ "entropy": 0.4786584210395813,
+ "epoch": 3.1542957334891875,
+ "grad_norm": 0.6474543809890747,
+ "learning_rate": 0.00022816847913292434,
+ "loss": 0.4224985122680664,
+ "mean_token_accuracy": 0.8628773537278175,
+ "num_tokens": 4180560.0,
+ "step": 1350
+ },
+ {
+ "entropy": 0.49172899857163427,
+ "epoch": 3.2711864406779663,
+ "grad_norm": 0.549326479434967,
+ "learning_rate": 0.00022442078300588989,
+ "loss": 0.43718509674072265,
+ "mean_token_accuracy": 0.8583108323812485,
+ "num_tokens": 4329954.0,
+ "step": 1400
+ },
+ {
+ "entropy": 0.4829080794751644,
+ "epoch": 3.3880771478667446,
+ "grad_norm": 0.5274888873100281,
+ "learning_rate": 0.00022051912210418922,
+ "loss": 0.4315507125854492,
+ "mean_token_accuracy": 0.8612637943029404,
+ "num_tokens": 4484354.0,
+ "step": 1450
+ },
+ {
+ "entropy": 0.4855460473895073,
+ "epoch": 3.504967855055523,
+ "grad_norm": 0.5846306681632996,
+ "learning_rate": 0.00021646998361658985,
+ "loss": 0.43724521636962893,
+ "mean_token_accuracy": 0.8589270997047425,
+ "num_tokens": 4638919.0,
+ "step": 1500
+ },
+ {
+ "entropy": 0.4905838692188263,
+ "epoch": 3.6218585622443014,
+ "grad_norm": 0.5149123072624207,
+ "learning_rate": 0.00021228009993895495,
+ "loss": 0.44197372436523436,
+ "mean_token_accuracy": 0.8584243017435074,
+ "num_tokens": 4795540.0,
+ "step": 1550
+ },
+ {
+ "entropy": 0.4994237454235554,
+ "epoch": 3.73874926943308,
+ "grad_norm": 0.5417225956916809,
+ "learning_rate": 0.0002079564374804665,
+ "loss": 0.4491108322143555,
+ "mean_token_accuracy": 0.8556393891572952,
+ "num_tokens": 4951177.0,
+ "step": 1600
+ },
+ {
+ "entropy": 0.49903531357645986,
+ "epoch": 3.8556399766218585,
+ "grad_norm": 0.5086734890937805,
+ "learning_rate": 0.00020350618508076015,
+ "loss": 0.4518918228149414,
+ "mean_token_accuracy": 0.8549707370996475,
+ "num_tokens": 5104814.0,
+ "step": 1650
+ },
+ {
+ "entropy": 0.4859523306787014,
+ "epoch": 3.972530683810637,
+ "grad_norm": 0.5446717739105225,
+ "learning_rate": 0.00019893674205723094,
+ "loss": 0.4410313415527344,
+ "mean_token_accuracy": 0.8587654489278793,
+ "num_tokens": 5265185.0,
+ "step": 1700
+ },
+ {
+ "epoch": 4.0,
+ "eval_entropy": 0.6577693150415049,
+ "eval_loss": 0.7380754947662354,
+ "eval_mean_token_accuracy": 0.8039549565934515,
+ "eval_num_tokens": 5302572.0,
+ "eval_runtime": 36.5757,
+ "eval_samples_per_second": 33.656,
+ "eval_steps_per_second": 4.21,
+ "step": 1712
+ },
+ {
+ "entropy": 0.41945636961328325,
+ "epoch": 4.088836937463472,
+ "grad_norm": 0.5925881266593933,
+ "learning_rate": 0.00019425570590238332,
+ "loss": 0.3575031280517578,
+ "mean_token_accuracy": 0.8812823918596584,
+ "num_tokens": 5419002.0,
+ "step": 1750
+ },
+ {
+ "entropy": 0.40106018364429474,
+ "epoch": 4.20572764465225,
+ "grad_norm": 0.7094829082489014,
+ "learning_rate": 0.0001894708596516806,
+ "loss": 0.3372145080566406,
+ "mean_token_accuracy": 0.8856011089682579,
+ "num_tokens": 5572969.0,
+ "step": 1800
+ },
+ {
+ "entropy": 0.4121378688514233,
+ "epoch": 4.322618351841029,
+ "grad_norm": 0.6077721118927002,
+ "learning_rate": 0.00018459015894289695,
+ "loss": 0.34956165313720705,
+ "mean_token_accuracy": 0.8818695649504662,
+ "num_tokens": 5733805.0,
+ "step": 1850
+ },
+ {
+ "entropy": 0.40683484479784965,
+ "epoch": 4.439509059029807,
+ "grad_norm": 0.5614883303642273,
+ "learning_rate": 0.00017962171878848807,
+ "loss": 0.34749691009521483,
+ "mean_token_accuracy": 0.882930768430233,
+ "num_tokens": 5888694.0,
+ "step": 1900
+ },
+ {
+ "entropy": 0.4100585526227951,
+ "epoch": 4.556399766218585,
+ "grad_norm": 0.5506443977355957,
+ "learning_rate": 0.00017457380008297383,
+ "loss": 0.3501173782348633,
+ "mean_token_accuracy": 0.8827524068951607,
+ "num_tokens": 6042655.0,
+ "step": 1950
+ },
+ {
+ "entropy": 0.41173958137631417,
+ "epoch": 4.673290473407365,
+ "grad_norm": 0.5409566164016724,
+ "learning_rate": 0.00016945479586776672,
+ "loss": 0.3541391754150391,
+ "mean_token_accuracy": 0.8808356913924217,
+ "num_tokens": 6203980.0,
+ "step": 2000
+ },
+ {
+ "entropy": 0.4158783262968063,
+ "epoch": 4.790181180596143,
+ "grad_norm": 0.572618842124939,
+ "learning_rate": 0.00016427321737628295,
+ "loss": 0.3536404800415039,
+ "mean_token_accuracy": 0.880456589460373,
+ "num_tokens": 6354604.0,
+ "step": 2050
+ },
+ {
+ "entropy": 0.408166281580925,
+ "epoch": 4.907071887784921,
+ "grad_norm": 0.5987762808799744,
+ "learning_rate": 0.00015903767988253897,
+ "loss": 0.3518599319458008,
+ "mean_token_accuracy": 0.8815453574061394,
+ "num_tokens": 6507484.0,
+ "step": 2100
+ },
+ {
+ "epoch": 5.0,
+ "eval_entropy": 0.6047254425364655,
+ "eval_loss": 0.7816920876502991,
+ "eval_mean_token_accuracy": 0.7987685733801358,
+ "eval_num_tokens": 6628215.0,
+ "eval_runtime": 36.9374,
+ "eval_samples_per_second": 33.327,
+ "eval_steps_per_second": 4.169,
+ "step": 2140
+ },
+ {
+ "entropy": 0.38985791001188097,
+ "epoch": 5.023378141437756,
+ "grad_norm": 0.7454882860183716,
+ "learning_rate": 0.00015375688837676246,
+ "loss": 0.32916854858398437,
+ "mean_token_accuracy": 0.8886270373310875,
+ "num_tokens": 6662015.0,
+ "step": 2150
+ },
+ {
+ "entropy": 0.30170531339943407,
+ "epoch": 5.140268848626534,
+ "grad_norm": 0.5329811573028564,
+ "learning_rate": 0.0001484396230918344,
+ "loss": 0.23112209320068358,
+ "mean_token_accuracy": 0.9198204749822616,
+ "num_tokens": 6817989.0,
+ "step": 2200
+ },
+ {
+ "entropy": 0.3075008810311556,
+ "epoch": 5.257159555815313,
+ "grad_norm": 0.6896093487739563,
+ "learning_rate": 0.00014309472490462715,
+ "loss": 0.2416652488708496,
+ "mean_token_accuracy": 0.9158366405963898,
+ "num_tokens": 6977021.0,
+ "step": 2250
+ },
+ {
+ "entropy": 0.31153449676930905,
+ "epoch": 5.374050263004091,
+ "grad_norm": 0.6930833458900452,
+ "learning_rate": 0.00013773108063651161,
+ "loss": 0.24753889083862304,
+ "mean_token_accuracy": 0.9136614826321602,
+ "num_tokens": 7133360.0,
+ "step": 2300
+ },
+ {
+ "entropy": 0.3154346436262131,
+ "epoch": 5.490940970192869,
+ "grad_norm": 0.6476582884788513,
+ "learning_rate": 0.0001323576082774734,
+ "loss": 0.24872392654418946,
+ "mean_token_accuracy": 0.9135914054512978,
+ "num_tokens": 7282372.0,
+ "step": 2350
+ },
+ {
+ "entropy": 0.313641587048769,
+ "epoch": 5.607831677381649,
+ "grad_norm": 0.6953482627868652,
+ "learning_rate": 0.0001269832421584059,
+ "loss": 0.24919464111328124,
+ "mean_token_accuracy": 0.9131662029027939,
+ "num_tokens": 7435766.0,
+ "step": 2400
+ },
+ {
+ "entropy": 0.3163894522190094,
+ "epoch": 5.724722384570427,
+ "grad_norm": 0.6673940420150757,
+ "learning_rate": 0.00012161691809623384,
+ "loss": 0.25111726760864256,
+ "mean_token_accuracy": 0.913602537214756,
+ "num_tokens": 7578408.0,
+ "step": 2450
+ },
+ {
+ "entropy": 0.3183368914574385,
+ "epoch": 5.841613091759205,
+ "grad_norm": 0.7174487709999084,
+ "learning_rate": 0.00011626755853656576,
+ "loss": 0.2518378448486328,
+ "mean_token_accuracy": 0.912000589966774,
+ "num_tokens": 7734599.0,
+ "step": 2500
+ },
+ {
+ "entropy": 0.3077564483135939,
+ "epoch": 5.958503798947984,
+ "grad_norm": 0.6662200689315796,
+ "learning_rate": 0.00011094405771857891,
+ "loss": 0.24354087829589843,
+ "mean_token_accuracy": 0.9140791621804237,
+ "num_tokens": 7898028.0,
+ "step": 2550
+ },
+ {
+ "epoch": 6.0,
+ "eval_entropy": 0.5403795917699863,
+ "eval_loss": 0.8572548627853394,
+ "eval_mean_token_accuracy": 0.796143418395674,
+ "eval_num_tokens": 7953858.0,
+ "eval_runtime": 36.7176,
+ "eval_samples_per_second": 33.526,
+ "eval_steps_per_second": 4.194,
+ "step": 2568
+ },
+ {
+ "entropy": 0.25110363144071857,
+ "epoch": 6.074810052600818,
+ "grad_norm": 0.6466763019561768,
+ "learning_rate": 0.0001056552668868023,
+ "loss": 0.1814822769165039,
+ "mean_token_accuracy": 0.9370300626035911,
+ "num_tokens": 8052646.0,
+ "step": 2600
+ },
+ {
+ "entropy": 0.2063255251944065,
+ "epoch": 6.1917007597895966,
+ "grad_norm": 0.6950973272323608,
+ "learning_rate": 0.0001004099795743853,
+ "loss": 0.13960565567016603,
+ "mean_token_accuracy": 0.9511301657557487,
+ "num_tokens": 8212323.0,
+ "step": 2650
+ },
+ {
+ "entropy": 0.20775293290615082,
+ "epoch": 6.308591466978375,
+ "grad_norm": 0.660478413105011,
+ "learning_rate": 9.52169169823223e-05,
+ "loss": 0.14319146156311036,
+ "mean_token_accuracy": 0.9500676044821739,
+ "num_tokens": 8366715.0,
+ "step": 2700
+ },
+ {
+ "entropy": 0.20977566957473756,
+ "epoch": 6.425482174167154,
+ "grad_norm": 0.6141477823257446,
+ "learning_rate": 9.00847134789411e-05,
+ "loss": 0.14525352478027342,
+ "mean_token_accuracy": 0.9490846493840217,
+ "num_tokens": 8515955.0,
+ "step": 2750
+ },
+ {
+ "entropy": 0.2143413795530796,
+ "epoch": 6.5423728813559325,
+ "grad_norm": 0.612184464931488,
+ "learning_rate": 8.502190224376661e-05,
+ "loss": 0.14751774787902833,
+ "mean_token_accuracy": 0.9478235357999801,
+ "num_tokens": 8672361.0,
+ "step": 2800
+ },
+ {
+ "entropy": 0.21027401253581046,
+ "epoch": 6.659263588544711,
+ "grad_norm": 0.644019365310669,
+ "learning_rate": 8.003690107962787e-05,
+ "loss": 0.1495517921447754,
+ "mean_token_accuracy": 0.9470956793427467,
+ "num_tokens": 8827819.0,
+ "step": 2850
+ },
+ {
+ "entropy": 0.2109992092102766,
+ "epoch": 6.776154295733489,
+ "grad_norm": 0.5456500053405762,
+ "learning_rate": 7.513799841659927e-05,
+ "loss": 0.14893126487731934,
+ "mean_token_accuracy": 0.9479587042331695,
+ "num_tokens": 8979535.0,
+ "step": 2900
+ },
+ {
+ "entropy": 0.20949276342988013,
+ "epoch": 6.893045002922268,
+ "grad_norm": 0.6803008913993835,
+ "learning_rate": 7.033333953104615e-05,
+ "loss": 0.1461741542816162,
+ "mean_token_accuracy": 0.9480843073129654,
+ "num_tokens": 9134134.0,
+ "step": 2950
+ },
+ {
+ "epoch": 7.0,
+ "eval_entropy": 0.4759114793949313,
+ "eval_loss": 0.9827883839607239,
+ "eval_mean_token_accuracy": 0.7945680889216337,
+ "eval_num_tokens": 9279501.0,
+ "eval_runtime": 36.4188,
+ "eval_samples_per_second": 33.801,
+ "eval_steps_per_second": 4.229,
+ "step": 2996
+ },
+ {
+ "entropy": 0.20382985060838,
+ "epoch": 7.009351256575102,
+ "grad_norm": 0.42079976201057434,
+ "learning_rate": 6.563091300268857e-05,
+ "loss": 0.13976585388183593,
+ "mean_token_accuracy": 0.9511888485458029,
+ "num_tokens": 9290416.0,
+ "step": 3000
+ },
+ {
+ "entropy": 0.14199006125330926,
+ "epoch": 7.1262419637638805,
+ "grad_norm": 0.6262232065200806,
+ "learning_rate": 6.103853743220019e-05,
+ "loss": 0.08045654296875,
+ "mean_token_accuracy": 0.9727814987301826,
+ "num_tokens": 9444705.0,
+ "step": 3050
+ },
+ {
+ "entropy": 0.13681559398770332,
+ "epoch": 7.243132670952659,
+ "grad_norm": 0.6460694074630737,
+ "learning_rate": 5.6563848441427006e-05,
+ "loss": 0.0795314359664917,
+ "mean_token_accuracy": 0.973130469918251,
+ "num_tokens": 9601476.0,
+ "step": 3100
+ },
+ {
+ "entropy": 0.13472390301525594,
+ "epoch": 7.360023378141438,
+ "grad_norm": 0.4053943157196045,
+ "learning_rate": 5.221428597784018e-05,
+ "loss": 0.07887161254882813,
+ "mean_token_accuracy": 0.9736681738495827,
+ "num_tokens": 9757913.0,
+ "step": 3150
+ },
+ {
+ "entropy": 0.14234945647418498,
+ "epoch": 7.4769140853302165,
+ "grad_norm": 0.46239086985588074,
+ "learning_rate": 4.7997081944331205e-05,
+ "loss": 0.08276874542236329,
+ "mean_token_accuracy": 0.9716356429457664,
+ "num_tokens": 9905790.0,
+ "step": 3200
+ },
+ {
+ "entropy": 0.13719632107764482,
+ "epoch": 7.593804792518995,
+ "grad_norm": 0.49838459491729736,
+ "learning_rate": 4.3919248174917856e-05,
+ "loss": 0.08003978729248047,
+ "mean_token_accuracy": 0.9731476292014122,
+ "num_tokens": 10064234.0,
+ "step": 3250
+ },
+ {
+ "entropy": 0.13751005962491036,
+ "epoch": 7.710695499707773,
+ "grad_norm": 0.4959999620914459,
+ "learning_rate": 3.9987564776352335e-05,
+ "loss": 0.07964549541473388,
+ "mean_token_accuracy": 0.9722868010401726,
+ "num_tokens": 10221089.0,
+ "step": 3300
+ },
+ {
+ "entropy": 0.13423183601349592,
+ "epoch": 7.827586206896552,
+ "grad_norm": 0.4889160692691803,
+ "learning_rate": 3.620856885501646e-05,
+ "loss": 0.07869738101959228,
+ "mean_token_accuracy": 0.973022837638855,
+ "num_tokens": 10379655.0,
+ "step": 3350
+ },
+ {
+ "entropy": 0.13587324120104313,
+ "epoch": 7.94447691408533,
+ "grad_norm": 0.4851946234703064,
+ "learning_rate": 3.258854364784726e-05,
+ "loss": 0.08004307746887207,
+ "mean_token_accuracy": 0.9723974686861038,
+ "num_tokens": 10531301.0,
+ "step": 3400
+ },
+ {
+ "epoch": 8.0,
+ "eval_entropy": 0.4391631047640528,
+ "eval_loss": 1.1321114301681519,
+ "eval_mean_token_accuracy": 0.7924294758152652,
+ "eval_num_tokens": 10605144.0,
+ "eval_runtime": 37.126,
+ "eval_samples_per_second": 33.157,
+ "eval_steps_per_second": 4.148,
+ "step": 3424
+ },
+ {
+ "entropy": 0.12316639324528487,
+ "epoch": 8.060783167738165,
+ "grad_norm": 0.3510015606880188,
+ "learning_rate": 2.9133508075364238e-05,
+ "loss": 0.06677456378936768,
+ "mean_token_accuracy": 0.9770565955483135,
+ "num_tokens": 10686079.0,
+ "step": 3450
+ },
+ {
+ "entropy": 0.11033655313774943,
+ "epoch": 8.177673874926944,
+ "grad_norm": 0.27871957421302795,
+ "learning_rate": 2.5849206734169147e-05,
+ "loss": 0.05629729747772217,
+ "mean_token_accuracy": 0.9804861381649971,
+ "num_tokens": 10838236.0,
+ "step": 3500
+ },
+ {
+ "entropy": 0.10826377695426345,
+ "epoch": 8.294564582115722,
+ "grad_norm": 0.44871190190315247,
+ "learning_rate": 2.2741100345556444e-05,
+ "loss": 0.05603146076202393,
+ "mean_token_accuracy": 0.9803140532970428,
+ "num_tokens": 10989653.0,
+ "step": 3550
+ },
+ {
+ "entropy": 0.10614268690347671,
+ "epoch": 8.4114552893045,
+ "grad_norm": 0.25647690892219543,
+ "learning_rate": 1.9814356676115743e-05,
+ "loss": 0.0554099178314209,
+ "mean_token_accuracy": 0.9802037307620048,
+ "num_tokens": 11144096.0,
+ "step": 3600
+ },
+ {
+ "entropy": 0.10369085857644678,
+ "epoch": 8.528345996493279,
+ "grad_norm": 0.32260316610336304,
+ "learning_rate": 1.7073841945422747e-05,
+ "loss": 0.05409110546112061,
+ "mean_token_accuracy": 0.980969468653202,
+ "num_tokens": 11303465.0,
+ "step": 3650
+ },
+ {
+ "entropy": 0.10663174549117685,
+ "epoch": 8.645236703682057,
+ "grad_norm": 0.33514541387557983,
+ "learning_rate": 1.4524112735104158e-05,
+ "loss": 0.056715693473815915,
+ "mean_token_accuracy": 0.9798875021934509,
+ "num_tokens": 11455533.0,
+ "step": 3700
+ },
+ {
+ "entropy": 0.1048200499266386,
+ "epoch": 8.762127410870836,
+ "grad_norm": 0.3340476453304291,
+ "learning_rate": 1.2169408412729229e-05,
+ "loss": 0.05345354557037354,
+ "mean_token_accuracy": 0.9803790295124054,
+ "num_tokens": 11618619.0,
+ "step": 3750
+ },
+ {
+ "entropy": 0.10449992345646024,
+ "epoch": 8.879018118059614,
+ "grad_norm": 0.30406567454338074,
+ "learning_rate": 1.0013644083125266e-05,
+ "loss": 0.054877762794494626,
+ "mean_token_accuracy": 0.9807043489813805,
+ "num_tokens": 11774627.0,
+ "step": 3800
+ },
+ {
+ "entropy": 0.10615571580827236,
+ "epoch": 8.995908825248392,
+ "grad_norm": 0.36648011207580566,
+ "learning_rate": 8.060404078835946e-06,
+ "loss": 0.056678996086120606,
+ "mean_token_accuracy": 0.9797374567389489,
+ "num_tokens": 11925408.0,
+ "step": 3850
+ },
+ {
+ "epoch": 9.0,
+ "eval_entropy": 0.4037178186820699,
+ "eval_loss": 1.254728078842163,
+ "eval_mean_token_accuracy": 0.7945115701718763,
+ "eval_num_tokens": 11930787.0,
+ "eval_runtime": 36.9056,
+ "eval_samples_per_second": 33.355,
+ "eval_steps_per_second": 4.173,
+ "step": 3852
+ },
+ {
+ "entropy": 0.09968331250263818,
+ "epoch": 9.112215078901228,
+ "grad_norm": 0.21597710251808167,
+ "learning_rate": 6.312936000545962e-06,
+ "loss": 0.04916244506835937,
+ "mean_token_accuracy": 0.9826284425342502,
+ "num_tokens": 12078366.0,
+ "step": 3900
+ },
+ {
+ "entropy": 0.1036025501601398,
+ "epoch": 9.229105786090006,
+ "grad_norm": 0.1738789677619934,
+ "learning_rate": 4.774145317381034e-06,
+ "loss": 0.05066513061523437,
+ "mean_token_accuracy": 0.9816445669531823,
+ "num_tokens": 12229118.0,
+ "step": 3950
+ },
+ {
+ "entropy": 0.09634713580831886,
+ "epoch": 9.345996493278784,
+ "grad_norm": 0.15528039634227753,
+ "learning_rate": 3.4465905360609734e-06,
+ "loss": 0.048097033500671384,
+ "mean_token_accuracy": 0.9828672608733178,
+ "num_tokens": 12385835.0,
+ "step": 4000
+ },
+ {
+ "entropy": 0.09529286917299032,
+ "epoch": 9.462887200467563,
+ "grad_norm": 0.3298656940460205,
+ "learning_rate": 2.3324789469378376e-06,
+ "loss": 0.04854735374450683,
+ "mean_token_accuracy": 0.9828567364811898,
+ "num_tokens": 12543991.0,
+ "step": 4050
+ },
+ {
+ "entropy": 0.09487062746658921,
+ "epoch": 9.579777907656341,
+ "grad_norm": 0.1514902114868164,
+ "learning_rate": 1.4336629539926327e-06,
+ "loss": 0.04930572509765625,
+ "mean_token_accuracy": 0.982597168982029,
+ "num_tokens": 12701866.0,
+ "step": 4100
+ },
+ {
+ "entropy": 0.09462659215554595,
+ "epoch": 9.69666861484512,
+ "grad_norm": 0.20895801484584808,
+ "learning_rate": 7.516369948918152e-07,
+ "loss": 0.048983674049377444,
+ "mean_token_accuracy": 0.9824120745062828,
+ "num_tokens": 12858675.0,
+ "step": 4150
+ },
+ {
+ "entropy": 0.09866151874884962,
+ "epoch": 9.813559322033898,
+ "grad_norm": 0.19578123092651367,
+ "learning_rate": 2.8753505622533914e-07,
+ "loss": 0.050036773681640626,
+ "mean_token_accuracy": 0.9818735978007317,
+ "num_tokens": 13010982.0,
+ "step": 4200
+ },
+ {
+ "entropy": 0.10088366318494081,
+ "epoch": 9.930450029222676,
+ "grad_norm": 0.26500314474105835,
+ "learning_rate": 4.2128788056935204e-08,
+ "loss": 0.0506698226928711,
+ "mean_token_accuracy": 0.9811591205000877,
+ "num_tokens": 13161041.0,
+ "step": 4250
+ },
+ {
+ "epoch": 10.0,
+ "eval_entropy": 0.39390789359420925,
+ "eval_loss": 1.3178749084472656,
+ "eval_mean_token_accuracy": 0.7940253568934156,
+ "eval_num_tokens": 13256430.0,
+ "eval_runtime": 36.3673,
+ "eval_samples_per_second": 33.849,
+ "eval_steps_per_second": 4.235,
+ "step": 4280
+ }
+ ],
+ "logging_steps": 50,
+ "max_steps": 4280,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 10,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": true
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 2.046123177338496e+17,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-856/README.md b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-856/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-856/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-856/adapter_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-856/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9e45544de727016be134f25e606b7b2b87c67549
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-856/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.050044417519510255,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "v_proj",
+ "o_proj",
+ "q_proj",
+ "gate_proj",
+ "k_proj",
+ "down_proj",
+ "up_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-856/chat_template.jinja b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-856/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-856/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-856/tokenizer_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-856/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-856/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-856/trainer_state.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-856/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..da46d88888cf4e6160b9e71d92c4f14b1cea021f
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test2/checkpoint-856/trainer_state.json
@@ -0,0 +1,226 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 2.0,
+ "eval_steps": 500,
+ "global_step": 856,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 1.8448095142841339,
+ "epoch": 0.11689070718877849,
+ "grad_norm": 1.4738068580627441,
+ "learning_rate": 3.0183098859435753e-05,
+ "loss": 1.6948321533203126,
+ "mean_token_accuracy": 0.6315939916670322,
+ "num_tokens": 153689.0,
+ "step": 50
+ },
+ {
+ "entropy": 0.9560942512750625,
+ "epoch": 0.23378141437755698,
+ "grad_norm": 1.378090739250183,
+ "learning_rate": 6.098217932824775e-05,
+ "loss": 0.8888477325439453,
+ "mean_token_accuracy": 0.7586692428588867,
+ "num_tokens": 310548.0,
+ "step": 100
+ },
+ {
+ "entropy": 0.848925698697567,
+ "epoch": 0.3506721215663355,
+ "grad_norm": 0.959621250629425,
+ "learning_rate": 9.178125979705974e-05,
+ "loss": 0.7850731658935547,
+ "mean_token_accuracy": 0.7774798944592476,
+ "num_tokens": 468560.0,
+ "step": 150
+ },
+ {
+ "entropy": 0.811761694252491,
+ "epoch": 0.46756282875511396,
+ "grad_norm": 0.8867514133453369,
+ "learning_rate": 0.00012258034026587175,
+ "loss": 0.7446884155273438,
+ "mean_token_accuracy": 0.7892421445250511,
+ "num_tokens": 619552.0,
+ "step": 200
+ },
+ {
+ "entropy": 0.7709847274422645,
+ "epoch": 0.5844535359438925,
+ "grad_norm": 0.5822142362594604,
+ "learning_rate": 0.00015337942073468373,
+ "loss": 0.7130480194091797,
+ "mean_token_accuracy": 0.7964023387432099,
+ "num_tokens": 771972.0,
+ "step": 250
+ },
+ {
+ "entropy": 0.7401632118225098,
+ "epoch": 0.701344243132671,
+ "grad_norm": 0.7557511925697327,
+ "learning_rate": 0.00018417850120349574,
+ "loss": 0.6857429504394531,
+ "mean_token_accuracy": 0.8041065946221352,
+ "num_tokens": 928088.0,
+ "step": 300
+ },
+ {
+ "entropy": 0.7295274111628532,
+ "epoch": 0.8182349503214494,
+ "grad_norm": 0.6769705414772034,
+ "learning_rate": 0.00021497758167230772,
+ "loss": 0.6801216125488281,
+ "mean_token_accuracy": 0.8037989619374275,
+ "num_tokens": 1086836.0,
+ "step": 350
+ },
+ {
+ "entropy": 0.710282272696495,
+ "epoch": 0.9351256575102279,
+ "grad_norm": 0.832009494304657,
+ "learning_rate": 0.00024577666214111967,
+ "loss": 0.6587177276611328,
+ "mean_token_accuracy": 0.8084024578332901,
+ "num_tokens": 1241227.0,
+ "step": 400
+ },
+ {
+ "epoch": 1.0,
+ "eval_entropy": 0.848237401672772,
+ "eval_loss": 0.7832363247871399,
+ "eval_mean_token_accuracy": 0.7817642971292719,
+ "eval_num_tokens": 1325643.0,
+ "eval_runtime": 36.8595,
+ "eval_samples_per_second": 33.397,
+ "eval_steps_per_second": 4.178,
+ "step": 428
+ },
+ {
+ "entropy": 0.7039205249230466,
+ "epoch": 1.0514319111630626,
+ "grad_norm": 0.6213153600692749,
+ "learning_rate": 0.00026362079548206905,
+ "loss": 0.6527813720703125,
+ "mean_token_accuracy": 0.8083000521564004,
+ "num_tokens": 1393922.0,
+ "step": 450
+ },
+ {
+ "entropy": 0.6749556151032448,
+ "epoch": 1.1683226183518411,
+ "grad_norm": 0.6444424390792847,
+ "learning_rate": 0.0002634191889632654,
+ "loss": 0.627038345336914,
+ "mean_token_accuracy": 0.8141268998384475,
+ "num_tokens": 1551442.0,
+ "step": 500
+ },
+ {
+ "entropy": 0.6727319830656051,
+ "epoch": 1.2852133255406195,
+ "grad_norm": 0.6360110640525818,
+ "learning_rate": 0.00026299877604786883,
+ "loss": 0.6232248687744141,
+ "mean_token_accuracy": 0.8153417313098907,
+ "num_tokens": 1706697.0,
+ "step": 550
+ },
+ {
+ "entropy": 0.6600116823613643,
+ "epoch": 1.4021040327293979,
+ "grad_norm": 0.46475180983543396,
+ "learning_rate": 0.00026236025574535523,
+ "loss": 0.6124750518798828,
+ "mean_token_accuracy": 0.8175293403863907,
+ "num_tokens": 1865246.0,
+ "step": 600
+ },
+ {
+ "entropy": 0.6494863349199295,
+ "epoch": 1.5189947399181765,
+ "grad_norm": 0.6132755875587463,
+ "learning_rate": 0.00026150468970660587,
+ "loss": 0.5997534942626953,
+ "mean_token_accuracy": 0.822050162255764,
+ "num_tokens": 2022060.0,
+ "step": 650
+ },
+ {
+ "entropy": 0.6511707258224487,
+ "epoch": 1.635885447106955,
+ "grad_norm": 0.570785641670227,
+ "learning_rate": 0.00026043350045872797,
+ "loss": 0.6059062576293945,
+ "mean_token_accuracy": 0.820236160159111,
+ "num_tokens": 2178909.0,
+ "step": 700
+ },
+ {
+ "entropy": 0.6550412529706955,
+ "epoch": 1.7527761542957334,
+ "grad_norm": 0.5398300290107727,
+ "learning_rate": 0.0002591484690398568,
+ "loss": 0.6031087875366211,
+ "mean_token_accuracy": 0.8206016793847084,
+ "num_tokens": 2329314.0,
+ "step": 750
+ },
+ {
+ "entropy": 0.6509851542115211,
+ "epoch": 1.869666861484512,
+ "grad_norm": 0.6913411617279053,
+ "learning_rate": 0.0002576517320378704,
+ "loss": 0.6023480224609375,
+ "mean_token_accuracy": 0.822216115295887,
+ "num_tokens": 2478559.0,
+ "step": 800
+ },
+ {
+ "entropy": 0.6560930109024048,
+ "epoch": 1.9865575686732906,
+ "grad_norm": 0.5568020939826965,
+ "learning_rate": 0.0002559457780379408,
+ "loss": 0.6031641006469727,
+ "mean_token_accuracy": 0.8205807811021805,
+ "num_tokens": 2631937.0,
+ "step": 850
+ },
+ {
+ "epoch": 2.0,
+ "eval_entropy": 0.794637631866839,
+ "eval_loss": 0.7390629649162292,
+ "eval_mean_token_accuracy": 0.7983863295672776,
+ "eval_num_tokens": 2651286.0,
+ "eval_runtime": 36.7346,
+ "eval_samples_per_second": 33.511,
+ "eval_steps_per_second": 4.192,
+ "step": 856
+ }
+ ],
+ "logging_steps": 50,
+ "max_steps": 4280,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 10,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 4.089312382391808e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..07feb095907d8ca70cf64cbd9167ace15b4e4f39
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/trainer_state.json
@@ -0,0 +1,139 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.24554941682013506,
+ "eval_steps": 20,
+ "global_step": 100,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ },
+ {
+ "entropy": 0.879152987897396,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 1.067221760749817,
+ "learning_rate": 0.00021473293557345806,
+ "loss": 0.807522201538086,
+ "mean_token_accuracy": 0.772921597212553,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.8607830974033901,
+ "eval_loss": 0.8036007881164551,
+ "eval_mean_token_accuracy": 0.7755460752759661,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 57.7576,
+ "eval_samples_per_second": 24.17,
+ "eval_steps_per_second": 3.03,
+ "step": 100
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 4385149329799680.0,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e50a57dda9ec2ad3d24d3a78331251fbce7557fb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/trainer_state.json
@@ -0,0 +1,160 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.2946593001841621,
+ "eval_steps": 20,
+ "global_step": 120,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ },
+ {
+ "entropy": 0.879152987897396,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 1.067221760749817,
+ "learning_rate": 0.00021473293557345806,
+ "loss": 0.807522201538086,
+ "mean_token_accuracy": 0.772921597212553,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.8607830974033901,
+ "eval_loss": 0.8036007881164551,
+ "eval_mean_token_accuracy": 0.7755460752759661,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 57.7576,
+ "eval_samples_per_second": 24.17,
+ "eval_steps_per_second": 3.03,
+ "step": 100
+ },
+ {
+ "entropy": 0.8589029781520366,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 0.7390042543411255,
+ "learning_rate": 0.0002136721067543722,
+ "loss": 0.7829637050628662,
+ "mean_token_accuracy": 0.779672809690237,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8330987681661334,
+ "eval_loss": 0.7742031216621399,
+ "eval_mean_token_accuracy": 0.785119104044778,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 57.3803,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 120
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 5253668222392320.0,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..306528192ea7d90fc62e2c168252db6d1c48eb61
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/trainer_state.json
@@ -0,0 +1,181 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.3437691835481891,
+ "eval_steps": 20,
+ "global_step": 140,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ },
+ {
+ "entropy": 0.879152987897396,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 1.067221760749817,
+ "learning_rate": 0.00021473293557345806,
+ "loss": 0.807522201538086,
+ "mean_token_accuracy": 0.772921597212553,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.8607830974033901,
+ "eval_loss": 0.8036007881164551,
+ "eval_mean_token_accuracy": 0.7755460752759661,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 57.7576,
+ "eval_samples_per_second": 24.17,
+ "eval_steps_per_second": 3.03,
+ "step": 100
+ },
+ {
+ "entropy": 0.8589029781520366,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 0.7390042543411255,
+ "learning_rate": 0.0002136721067543722,
+ "loss": 0.7829637050628662,
+ "mean_token_accuracy": 0.779672809690237,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8330987681661334,
+ "eval_loss": 0.7742031216621399,
+ "eval_mean_token_accuracy": 0.785119104044778,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 57.3803,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 120
+ },
+ {
+ "entropy": 0.8094905942678452,
+ "epoch": 0.3437691835481891,
+ "grad_norm": 0.6055881381034851,
+ "learning_rate": 0.00021183381937537642,
+ "loss": 0.7396022796630859,
+ "mean_token_accuracy": 0.7867072127759457,
+ "num_tokens": 361815.0,
+ "step": 140
+ },
+ {
+ "epoch": 0.3437691835481891,
+ "eval_entropy": 0.7953300523757935,
+ "eval_loss": 0.7521983981132507,
+ "eval_mean_token_accuracy": 0.7905156462533134,
+ "eval_num_tokens": 361815.0,
+ "eval_runtime": 57.6345,
+ "eval_samples_per_second": 24.222,
+ "eval_steps_per_second": 3.036,
+ "step": 140
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 6118471451159040.0,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..1dc6737bd60fe2bc3a1de00ffc91ebb503692c2a
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/trainer_state.json
@@ -0,0 +1,202 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.3928790669122161,
+ "eval_steps": 20,
+ "global_step": 160,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ },
+ {
+ "entropy": 0.879152987897396,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 1.067221760749817,
+ "learning_rate": 0.00021473293557345806,
+ "loss": 0.807522201538086,
+ "mean_token_accuracy": 0.772921597212553,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.8607830974033901,
+ "eval_loss": 0.8036007881164551,
+ "eval_mean_token_accuracy": 0.7755460752759661,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 57.7576,
+ "eval_samples_per_second": 24.17,
+ "eval_steps_per_second": 3.03,
+ "step": 100
+ },
+ {
+ "entropy": 0.8589029781520366,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 0.7390042543411255,
+ "learning_rate": 0.0002136721067543722,
+ "loss": 0.7829637050628662,
+ "mean_token_accuracy": 0.779672809690237,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8330987681661334,
+ "eval_loss": 0.7742031216621399,
+ "eval_mean_token_accuracy": 0.785119104044778,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 57.3803,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 120
+ },
+ {
+ "entropy": 0.8094905942678452,
+ "epoch": 0.3437691835481891,
+ "grad_norm": 0.6055881381034851,
+ "learning_rate": 0.00021183381937537642,
+ "loss": 0.7396022796630859,
+ "mean_token_accuracy": 0.7867072127759457,
+ "num_tokens": 361815.0,
+ "step": 140
+ },
+ {
+ "epoch": 0.3437691835481891,
+ "eval_entropy": 0.7953300523757935,
+ "eval_loss": 0.7521983981132507,
+ "eval_mean_token_accuracy": 0.7905156462533134,
+ "eval_num_tokens": 361815.0,
+ "eval_runtime": 57.6345,
+ "eval_samples_per_second": 24.222,
+ "eval_steps_per_second": 3.036,
+ "step": 140
+ },
+ {
+ "entropy": 0.7955982603132725,
+ "epoch": 0.3928790669122161,
+ "grad_norm": 0.5252285599708557,
+ "learning_rate": 0.00020923153562932872,
+ "loss": 0.7219765186309814,
+ "mean_token_accuracy": 0.792796092480421,
+ "num_tokens": 417429.0,
+ "step": 160
+ },
+ {
+ "epoch": 0.3928790669122161,
+ "eval_entropy": 0.7623447510174343,
+ "eval_loss": 0.7298487424850464,
+ "eval_mean_token_accuracy": 0.7945844326700483,
+ "eval_num_tokens": 417429.0,
+ "eval_runtime": 57.9123,
+ "eval_samples_per_second": 24.105,
+ "eval_steps_per_second": 3.022,
+ "step": 160
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 7013736112611840.0,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..f1b2721eda1078756a54e1e0c1d6caf763d318ac
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/trainer_state.json
@@ -0,0 +1,223 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.4419889502762431,
+ "eval_steps": 20,
+ "global_step": 180,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ },
+ {
+ "entropy": 0.879152987897396,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 1.067221760749817,
+ "learning_rate": 0.00021473293557345806,
+ "loss": 0.807522201538086,
+ "mean_token_accuracy": 0.772921597212553,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.8607830974033901,
+ "eval_loss": 0.8036007881164551,
+ "eval_mean_token_accuracy": 0.7755460752759661,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 57.7576,
+ "eval_samples_per_second": 24.17,
+ "eval_steps_per_second": 3.03,
+ "step": 100
+ },
+ {
+ "entropy": 0.8589029781520366,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 0.7390042543411255,
+ "learning_rate": 0.0002136721067543722,
+ "loss": 0.7829637050628662,
+ "mean_token_accuracy": 0.779672809690237,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8330987681661334,
+ "eval_loss": 0.7742031216621399,
+ "eval_mean_token_accuracy": 0.785119104044778,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 57.3803,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 120
+ },
+ {
+ "entropy": 0.8094905942678452,
+ "epoch": 0.3437691835481891,
+ "grad_norm": 0.6055881381034851,
+ "learning_rate": 0.00021183381937537642,
+ "loss": 0.7396022796630859,
+ "mean_token_accuracy": 0.7867072127759457,
+ "num_tokens": 361815.0,
+ "step": 140
+ },
+ {
+ "epoch": 0.3437691835481891,
+ "eval_entropy": 0.7953300523757935,
+ "eval_loss": 0.7521983981132507,
+ "eval_mean_token_accuracy": 0.7905156462533134,
+ "eval_num_tokens": 361815.0,
+ "eval_runtime": 57.6345,
+ "eval_samples_per_second": 24.222,
+ "eval_steps_per_second": 3.036,
+ "step": 140
+ },
+ {
+ "entropy": 0.7955982603132725,
+ "epoch": 0.3928790669122161,
+ "grad_norm": 0.5252285599708557,
+ "learning_rate": 0.00020923153562932872,
+ "loss": 0.7219765186309814,
+ "mean_token_accuracy": 0.792796092480421,
+ "num_tokens": 417429.0,
+ "step": 160
+ },
+ {
+ "epoch": 0.3928790669122161,
+ "eval_entropy": 0.7623447510174343,
+ "eval_loss": 0.7298487424850464,
+ "eval_mean_token_accuracy": 0.7945844326700483,
+ "eval_num_tokens": 417429.0,
+ "eval_runtime": 57.9123,
+ "eval_samples_per_second": 24.105,
+ "eval_steps_per_second": 3.022,
+ "step": 160
+ },
+ {
+ "entropy": 0.7903043553233147,
+ "epoch": 0.4419889502762431,
+ "grad_norm": 0.6948235034942627,
+ "learning_rate": 0.00020588431262670093,
+ "loss": 0.7225021839141845,
+ "mean_token_accuracy": 0.791443546116352,
+ "num_tokens": 467817.0,
+ "step": 180
+ },
+ {
+ "epoch": 0.4419889502762431,
+ "eval_entropy": 0.790012172971453,
+ "eval_loss": 0.720366895198822,
+ "eval_mean_token_accuracy": 0.7943330574035644,
+ "eval_num_tokens": 467817.0,
+ "eval_runtime": 57.4578,
+ "eval_samples_per_second": 24.296,
+ "eval_steps_per_second": 3.046,
+ "step": 180
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 7814847254853120.0,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b2b6a96d0d8e21cecf76e6febd5c2d0685fb7a1b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/trainer_state.json
@@ -0,0 +1,55 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.04910988336402701,
+ "eval_steps": 20,
+ "global_step": 20,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 882154677765120.0,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..15d8b96e6d87d3f41a68155a2c10fa8f33753798
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/trainer_state.json
@@ -0,0 +1,244 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.4910988336402701,
+ "eval_steps": 20,
+ "global_step": 200,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ },
+ {
+ "entropy": 0.879152987897396,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 1.067221760749817,
+ "learning_rate": 0.00021473293557345806,
+ "loss": 0.807522201538086,
+ "mean_token_accuracy": 0.772921597212553,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.8607830974033901,
+ "eval_loss": 0.8036007881164551,
+ "eval_mean_token_accuracy": 0.7755460752759661,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 57.7576,
+ "eval_samples_per_second": 24.17,
+ "eval_steps_per_second": 3.03,
+ "step": 100
+ },
+ {
+ "entropy": 0.8589029781520366,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 0.7390042543411255,
+ "learning_rate": 0.0002136721067543722,
+ "loss": 0.7829637050628662,
+ "mean_token_accuracy": 0.779672809690237,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8330987681661334,
+ "eval_loss": 0.7742031216621399,
+ "eval_mean_token_accuracy": 0.785119104044778,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 57.3803,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 120
+ },
+ {
+ "entropy": 0.8094905942678452,
+ "epoch": 0.3437691835481891,
+ "grad_norm": 0.6055881381034851,
+ "learning_rate": 0.00021183381937537642,
+ "loss": 0.7396022796630859,
+ "mean_token_accuracy": 0.7867072127759457,
+ "num_tokens": 361815.0,
+ "step": 140
+ },
+ {
+ "epoch": 0.3437691835481891,
+ "eval_entropy": 0.7953300523757935,
+ "eval_loss": 0.7521983981132507,
+ "eval_mean_token_accuracy": 0.7905156462533134,
+ "eval_num_tokens": 361815.0,
+ "eval_runtime": 57.6345,
+ "eval_samples_per_second": 24.222,
+ "eval_steps_per_second": 3.036,
+ "step": 140
+ },
+ {
+ "entropy": 0.7955982603132725,
+ "epoch": 0.3928790669122161,
+ "grad_norm": 0.5252285599708557,
+ "learning_rate": 0.00020923153562932872,
+ "loss": 0.7219765186309814,
+ "mean_token_accuracy": 0.792796092480421,
+ "num_tokens": 417429.0,
+ "step": 160
+ },
+ {
+ "epoch": 0.3928790669122161,
+ "eval_entropy": 0.7623447510174343,
+ "eval_loss": 0.7298487424850464,
+ "eval_mean_token_accuracy": 0.7945844326700483,
+ "eval_num_tokens": 417429.0,
+ "eval_runtime": 57.9123,
+ "eval_samples_per_second": 24.105,
+ "eval_steps_per_second": 3.022,
+ "step": 160
+ },
+ {
+ "entropy": 0.7903043553233147,
+ "epoch": 0.4419889502762431,
+ "grad_norm": 0.6948235034942627,
+ "learning_rate": 0.00020588431262670093,
+ "loss": 0.7225021839141845,
+ "mean_token_accuracy": 0.791443546116352,
+ "num_tokens": 467817.0,
+ "step": 180
+ },
+ {
+ "epoch": 0.4419889502762431,
+ "eval_entropy": 0.790012172971453,
+ "eval_loss": 0.720366895198822,
+ "eval_mean_token_accuracy": 0.7943330574035644,
+ "eval_num_tokens": 467817.0,
+ "eval_runtime": 57.4578,
+ "eval_samples_per_second": 24.296,
+ "eval_steps_per_second": 3.046,
+ "step": 180
+ },
+ {
+ "entropy": 0.7773297265172004,
+ "epoch": 0.4910988336402701,
+ "grad_norm": 0.6160358786582947,
+ "learning_rate": 0.00020181666283606295,
+ "loss": 0.7100065231323243,
+ "mean_token_accuracy": 0.7956355385482311,
+ "num_tokens": 513773.0,
+ "step": 200
+ },
+ {
+ "epoch": 0.4910988336402701,
+ "eval_entropy": 0.783927013874054,
+ "eval_loss": 0.7202737927436829,
+ "eval_mean_token_accuracy": 0.7964657868657793,
+ "eval_num_tokens": 513773.0,
+ "eval_runtime": 57.6663,
+ "eval_samples_per_second": 24.208,
+ "eval_steps_per_second": 3.035,
+ "step": 200
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 8603198947729920.0,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..563306e129a331cecf8a3732127e315cfaf81ff9
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/trainer_state.json
@@ -0,0 +1,265 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.5402087170042971,
+ "eval_steps": 20,
+ "global_step": 220,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ },
+ {
+ "entropy": 0.879152987897396,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 1.067221760749817,
+ "learning_rate": 0.00021473293557345806,
+ "loss": 0.807522201538086,
+ "mean_token_accuracy": 0.772921597212553,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.8607830974033901,
+ "eval_loss": 0.8036007881164551,
+ "eval_mean_token_accuracy": 0.7755460752759661,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 57.7576,
+ "eval_samples_per_second": 24.17,
+ "eval_steps_per_second": 3.03,
+ "step": 100
+ },
+ {
+ "entropy": 0.8589029781520366,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 0.7390042543411255,
+ "learning_rate": 0.0002136721067543722,
+ "loss": 0.7829637050628662,
+ "mean_token_accuracy": 0.779672809690237,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8330987681661334,
+ "eval_loss": 0.7742031216621399,
+ "eval_mean_token_accuracy": 0.785119104044778,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 57.3803,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 120
+ },
+ {
+ "entropy": 0.8094905942678452,
+ "epoch": 0.3437691835481891,
+ "grad_norm": 0.6055881381034851,
+ "learning_rate": 0.00021183381937537642,
+ "loss": 0.7396022796630859,
+ "mean_token_accuracy": 0.7867072127759457,
+ "num_tokens": 361815.0,
+ "step": 140
+ },
+ {
+ "epoch": 0.3437691835481891,
+ "eval_entropy": 0.7953300523757935,
+ "eval_loss": 0.7521983981132507,
+ "eval_mean_token_accuracy": 0.7905156462533134,
+ "eval_num_tokens": 361815.0,
+ "eval_runtime": 57.6345,
+ "eval_samples_per_second": 24.222,
+ "eval_steps_per_second": 3.036,
+ "step": 140
+ },
+ {
+ "entropy": 0.7955982603132725,
+ "epoch": 0.3928790669122161,
+ "grad_norm": 0.5252285599708557,
+ "learning_rate": 0.00020923153562932872,
+ "loss": 0.7219765186309814,
+ "mean_token_accuracy": 0.792796092480421,
+ "num_tokens": 417429.0,
+ "step": 160
+ },
+ {
+ "epoch": 0.3928790669122161,
+ "eval_entropy": 0.7623447510174343,
+ "eval_loss": 0.7298487424850464,
+ "eval_mean_token_accuracy": 0.7945844326700483,
+ "eval_num_tokens": 417429.0,
+ "eval_runtime": 57.9123,
+ "eval_samples_per_second": 24.105,
+ "eval_steps_per_second": 3.022,
+ "step": 160
+ },
+ {
+ "entropy": 0.7903043553233147,
+ "epoch": 0.4419889502762431,
+ "grad_norm": 0.6948235034942627,
+ "learning_rate": 0.00020588431262670093,
+ "loss": 0.7225021839141845,
+ "mean_token_accuracy": 0.791443546116352,
+ "num_tokens": 467817.0,
+ "step": 180
+ },
+ {
+ "epoch": 0.4419889502762431,
+ "eval_entropy": 0.790012172971453,
+ "eval_loss": 0.720366895198822,
+ "eval_mean_token_accuracy": 0.7943330574035644,
+ "eval_num_tokens": 467817.0,
+ "eval_runtime": 57.4578,
+ "eval_samples_per_second": 24.296,
+ "eval_steps_per_second": 3.046,
+ "step": 180
+ },
+ {
+ "entropy": 0.7773297265172004,
+ "epoch": 0.4910988336402701,
+ "grad_norm": 0.6160358786582947,
+ "learning_rate": 0.00020181666283606295,
+ "loss": 0.7100065231323243,
+ "mean_token_accuracy": 0.7956355385482311,
+ "num_tokens": 513773.0,
+ "step": 200
+ },
+ {
+ "epoch": 0.4910988336402701,
+ "eval_entropy": 0.783927013874054,
+ "eval_loss": 0.7202737927436829,
+ "eval_mean_token_accuracy": 0.7964657868657793,
+ "eval_num_tokens": 513773.0,
+ "eval_runtime": 57.6663,
+ "eval_samples_per_second": 24.208,
+ "eval_steps_per_second": 3.035,
+ "step": 200
+ },
+ {
+ "entropy": 0.7645334511995315,
+ "epoch": 0.5402087170042971,
+ "grad_norm": 0.6620849967002869,
+ "learning_rate": 0.00019705837457374996,
+ "loss": 0.7037524223327637,
+ "mean_token_accuracy": 0.7996554173529148,
+ "num_tokens": 565606.0,
+ "step": 220
+ },
+ {
+ "epoch": 0.5402087170042971,
+ "eval_entropy": 0.7768624458994184,
+ "eval_loss": 0.7051481604576111,
+ "eval_mean_token_accuracy": 0.7991346962111336,
+ "eval_num_tokens": 565606.0,
+ "eval_runtime": 57.7442,
+ "eval_samples_per_second": 24.176,
+ "eval_steps_per_second": 3.031,
+ "step": 220
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 9483916434769920.0,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c7a6bbf613b56a80bf6945be589d0ee01252bd42
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/trainer_state.json
@@ -0,0 +1,286 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.5893186003683242,
+ "eval_steps": 20,
+ "global_step": 240,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ },
+ {
+ "entropy": 0.879152987897396,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 1.067221760749817,
+ "learning_rate": 0.00021473293557345806,
+ "loss": 0.807522201538086,
+ "mean_token_accuracy": 0.772921597212553,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.8607830974033901,
+ "eval_loss": 0.8036007881164551,
+ "eval_mean_token_accuracy": 0.7755460752759661,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 57.7576,
+ "eval_samples_per_second": 24.17,
+ "eval_steps_per_second": 3.03,
+ "step": 100
+ },
+ {
+ "entropy": 0.8589029781520366,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 0.7390042543411255,
+ "learning_rate": 0.0002136721067543722,
+ "loss": 0.7829637050628662,
+ "mean_token_accuracy": 0.779672809690237,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8330987681661334,
+ "eval_loss": 0.7742031216621399,
+ "eval_mean_token_accuracy": 0.785119104044778,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 57.3803,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 120
+ },
+ {
+ "entropy": 0.8094905942678452,
+ "epoch": 0.3437691835481891,
+ "grad_norm": 0.6055881381034851,
+ "learning_rate": 0.00021183381937537642,
+ "loss": 0.7396022796630859,
+ "mean_token_accuracy": 0.7867072127759457,
+ "num_tokens": 361815.0,
+ "step": 140
+ },
+ {
+ "epoch": 0.3437691835481891,
+ "eval_entropy": 0.7953300523757935,
+ "eval_loss": 0.7521983981132507,
+ "eval_mean_token_accuracy": 0.7905156462533134,
+ "eval_num_tokens": 361815.0,
+ "eval_runtime": 57.6345,
+ "eval_samples_per_second": 24.222,
+ "eval_steps_per_second": 3.036,
+ "step": 140
+ },
+ {
+ "entropy": 0.7955982603132725,
+ "epoch": 0.3928790669122161,
+ "grad_norm": 0.5252285599708557,
+ "learning_rate": 0.00020923153562932872,
+ "loss": 0.7219765186309814,
+ "mean_token_accuracy": 0.792796092480421,
+ "num_tokens": 417429.0,
+ "step": 160
+ },
+ {
+ "epoch": 0.3928790669122161,
+ "eval_entropy": 0.7623447510174343,
+ "eval_loss": 0.7298487424850464,
+ "eval_mean_token_accuracy": 0.7945844326700483,
+ "eval_num_tokens": 417429.0,
+ "eval_runtime": 57.9123,
+ "eval_samples_per_second": 24.105,
+ "eval_steps_per_second": 3.022,
+ "step": 160
+ },
+ {
+ "entropy": 0.7903043553233147,
+ "epoch": 0.4419889502762431,
+ "grad_norm": 0.6948235034942627,
+ "learning_rate": 0.00020588431262670093,
+ "loss": 0.7225021839141845,
+ "mean_token_accuracy": 0.791443546116352,
+ "num_tokens": 467817.0,
+ "step": 180
+ },
+ {
+ "epoch": 0.4419889502762431,
+ "eval_entropy": 0.790012172971453,
+ "eval_loss": 0.720366895198822,
+ "eval_mean_token_accuracy": 0.7943330574035644,
+ "eval_num_tokens": 467817.0,
+ "eval_runtime": 57.4578,
+ "eval_samples_per_second": 24.296,
+ "eval_steps_per_second": 3.046,
+ "step": 180
+ },
+ {
+ "entropy": 0.7773297265172004,
+ "epoch": 0.4910988336402701,
+ "grad_norm": 0.6160358786582947,
+ "learning_rate": 0.00020181666283606295,
+ "loss": 0.7100065231323243,
+ "mean_token_accuracy": 0.7956355385482311,
+ "num_tokens": 513773.0,
+ "step": 200
+ },
+ {
+ "epoch": 0.4910988336402701,
+ "eval_entropy": 0.783927013874054,
+ "eval_loss": 0.7202737927436829,
+ "eval_mean_token_accuracy": 0.7964657868657793,
+ "eval_num_tokens": 513773.0,
+ "eval_runtime": 57.6663,
+ "eval_samples_per_second": 24.208,
+ "eval_steps_per_second": 3.035,
+ "step": 200
+ },
+ {
+ "entropy": 0.7645334511995315,
+ "epoch": 0.5402087170042971,
+ "grad_norm": 0.6620849967002869,
+ "learning_rate": 0.00019705837457374996,
+ "loss": 0.7037524223327637,
+ "mean_token_accuracy": 0.7996554173529148,
+ "num_tokens": 565606.0,
+ "step": 220
+ },
+ {
+ "epoch": 0.5402087170042971,
+ "eval_entropy": 0.7768624458994184,
+ "eval_loss": 0.7051481604576111,
+ "eval_mean_token_accuracy": 0.7991346962111336,
+ "eval_num_tokens": 565606.0,
+ "eval_runtime": 57.7442,
+ "eval_samples_per_second": 24.176,
+ "eval_steps_per_second": 3.031,
+ "step": 220
+ },
+ {
+ "entropy": 0.7731568738818169,
+ "epoch": 0.5893186003683242,
+ "grad_norm": 0.553688645362854,
+ "learning_rate": 0.00019164429385730688,
+ "loss": 0.7003726959228516,
+ "mean_token_accuracy": 0.7977605648338795,
+ "num_tokens": 616528.0,
+ "step": 240
+ },
+ {
+ "epoch": 0.5893186003683242,
+ "eval_entropy": 0.7570131056649344,
+ "eval_loss": 0.7042225003242493,
+ "eval_mean_token_accuracy": 0.8000479991095407,
+ "eval_num_tokens": 616528.0,
+ "eval_runtime": 57.8713,
+ "eval_samples_per_second": 24.122,
+ "eval_steps_per_second": 3.024,
+ "step": 240
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.030707618594048e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ad4c55943e70fd22f234252755c40e9227febdb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/trainer_state.json
@@ -0,0 +1,307 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.6384284837323512,
+ "eval_steps": 20,
+ "global_step": 260,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ },
+ {
+ "entropy": 0.879152987897396,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 1.067221760749817,
+ "learning_rate": 0.00021473293557345806,
+ "loss": 0.807522201538086,
+ "mean_token_accuracy": 0.772921597212553,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.8607830974033901,
+ "eval_loss": 0.8036007881164551,
+ "eval_mean_token_accuracy": 0.7755460752759661,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 57.7576,
+ "eval_samples_per_second": 24.17,
+ "eval_steps_per_second": 3.03,
+ "step": 100
+ },
+ {
+ "entropy": 0.8589029781520366,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 0.7390042543411255,
+ "learning_rate": 0.0002136721067543722,
+ "loss": 0.7829637050628662,
+ "mean_token_accuracy": 0.779672809690237,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8330987681661334,
+ "eval_loss": 0.7742031216621399,
+ "eval_mean_token_accuracy": 0.785119104044778,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 57.3803,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 120
+ },
+ {
+ "entropy": 0.8094905942678452,
+ "epoch": 0.3437691835481891,
+ "grad_norm": 0.6055881381034851,
+ "learning_rate": 0.00021183381937537642,
+ "loss": 0.7396022796630859,
+ "mean_token_accuracy": 0.7867072127759457,
+ "num_tokens": 361815.0,
+ "step": 140
+ },
+ {
+ "epoch": 0.3437691835481891,
+ "eval_entropy": 0.7953300523757935,
+ "eval_loss": 0.7521983981132507,
+ "eval_mean_token_accuracy": 0.7905156462533134,
+ "eval_num_tokens": 361815.0,
+ "eval_runtime": 57.6345,
+ "eval_samples_per_second": 24.222,
+ "eval_steps_per_second": 3.036,
+ "step": 140
+ },
+ {
+ "entropy": 0.7955982603132725,
+ "epoch": 0.3928790669122161,
+ "grad_norm": 0.5252285599708557,
+ "learning_rate": 0.00020923153562932872,
+ "loss": 0.7219765186309814,
+ "mean_token_accuracy": 0.792796092480421,
+ "num_tokens": 417429.0,
+ "step": 160
+ },
+ {
+ "epoch": 0.3928790669122161,
+ "eval_entropy": 0.7623447510174343,
+ "eval_loss": 0.7298487424850464,
+ "eval_mean_token_accuracy": 0.7945844326700483,
+ "eval_num_tokens": 417429.0,
+ "eval_runtime": 57.9123,
+ "eval_samples_per_second": 24.105,
+ "eval_steps_per_second": 3.022,
+ "step": 160
+ },
+ {
+ "entropy": 0.7903043553233147,
+ "epoch": 0.4419889502762431,
+ "grad_norm": 0.6948235034942627,
+ "learning_rate": 0.00020588431262670093,
+ "loss": 0.7225021839141845,
+ "mean_token_accuracy": 0.791443546116352,
+ "num_tokens": 467817.0,
+ "step": 180
+ },
+ {
+ "epoch": 0.4419889502762431,
+ "eval_entropy": 0.790012172971453,
+ "eval_loss": 0.720366895198822,
+ "eval_mean_token_accuracy": 0.7943330574035644,
+ "eval_num_tokens": 467817.0,
+ "eval_runtime": 57.4578,
+ "eval_samples_per_second": 24.296,
+ "eval_steps_per_second": 3.046,
+ "step": 180
+ },
+ {
+ "entropy": 0.7773297265172004,
+ "epoch": 0.4910988336402701,
+ "grad_norm": 0.6160358786582947,
+ "learning_rate": 0.00020181666283606295,
+ "loss": 0.7100065231323243,
+ "mean_token_accuracy": 0.7956355385482311,
+ "num_tokens": 513773.0,
+ "step": 200
+ },
+ {
+ "epoch": 0.4910988336402701,
+ "eval_entropy": 0.783927013874054,
+ "eval_loss": 0.7202737927436829,
+ "eval_mean_token_accuracy": 0.7964657868657793,
+ "eval_num_tokens": 513773.0,
+ "eval_runtime": 57.6663,
+ "eval_samples_per_second": 24.208,
+ "eval_steps_per_second": 3.035,
+ "step": 200
+ },
+ {
+ "entropy": 0.7645334511995315,
+ "epoch": 0.5402087170042971,
+ "grad_norm": 0.6620849967002869,
+ "learning_rate": 0.00019705837457374996,
+ "loss": 0.7037524223327637,
+ "mean_token_accuracy": 0.7996554173529148,
+ "num_tokens": 565606.0,
+ "step": 220
+ },
+ {
+ "epoch": 0.5402087170042971,
+ "eval_entropy": 0.7768624458994184,
+ "eval_loss": 0.7051481604576111,
+ "eval_mean_token_accuracy": 0.7991346962111336,
+ "eval_num_tokens": 565606.0,
+ "eval_runtime": 57.7442,
+ "eval_samples_per_second": 24.176,
+ "eval_steps_per_second": 3.031,
+ "step": 220
+ },
+ {
+ "entropy": 0.7731568738818169,
+ "epoch": 0.5893186003683242,
+ "grad_norm": 0.553688645362854,
+ "learning_rate": 0.00019164429385730688,
+ "loss": 0.7003726959228516,
+ "mean_token_accuracy": 0.7977605648338795,
+ "num_tokens": 616528.0,
+ "step": 240
+ },
+ {
+ "epoch": 0.5893186003683242,
+ "eval_entropy": 0.7570131056649344,
+ "eval_loss": 0.7042225003242493,
+ "eval_mean_token_accuracy": 0.8000479991095407,
+ "eval_num_tokens": 616528.0,
+ "eval_runtime": 57.8713,
+ "eval_samples_per_second": 24.122,
+ "eval_steps_per_second": 3.024,
+ "step": 240
+ },
+ {
+ "entropy": 0.7343776792287826,
+ "epoch": 0.6384284837323512,
+ "grad_norm": 0.5576229095458984,
+ "learning_rate": 0.00018561406922024723,
+ "loss": 0.6735261440277099,
+ "mean_token_accuracy": 0.8059290319681167,
+ "num_tokens": 669440.0,
+ "step": 260
+ },
+ {
+ "epoch": 0.6384284837323512,
+ "eval_entropy": 0.7231997234480722,
+ "eval_loss": 0.6926313638687134,
+ "eval_mean_token_accuracy": 0.8031776939119611,
+ "eval_num_tokens": 669440.0,
+ "eval_runtime": 57.1692,
+ "eval_samples_per_second": 24.419,
+ "eval_steps_per_second": 3.061,
+ "step": 260
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.12234079602176e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ac80833bab7fe02d8d676d814f403fb9d04c92d8
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/trainer_state.json
@@ -0,0 +1,328 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.6875383670963782,
+ "eval_steps": 20,
+ "global_step": 280,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ },
+ {
+ "entropy": 0.879152987897396,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 1.067221760749817,
+ "learning_rate": 0.00021473293557345806,
+ "loss": 0.807522201538086,
+ "mean_token_accuracy": 0.772921597212553,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.8607830974033901,
+ "eval_loss": 0.8036007881164551,
+ "eval_mean_token_accuracy": 0.7755460752759661,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 57.7576,
+ "eval_samples_per_second": 24.17,
+ "eval_steps_per_second": 3.03,
+ "step": 100
+ },
+ {
+ "entropy": 0.8589029781520366,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 0.7390042543411255,
+ "learning_rate": 0.0002136721067543722,
+ "loss": 0.7829637050628662,
+ "mean_token_accuracy": 0.779672809690237,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8330987681661334,
+ "eval_loss": 0.7742031216621399,
+ "eval_mean_token_accuracy": 0.785119104044778,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 57.3803,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 120
+ },
+ {
+ "entropy": 0.8094905942678452,
+ "epoch": 0.3437691835481891,
+ "grad_norm": 0.6055881381034851,
+ "learning_rate": 0.00021183381937537642,
+ "loss": 0.7396022796630859,
+ "mean_token_accuracy": 0.7867072127759457,
+ "num_tokens": 361815.0,
+ "step": 140
+ },
+ {
+ "epoch": 0.3437691835481891,
+ "eval_entropy": 0.7953300523757935,
+ "eval_loss": 0.7521983981132507,
+ "eval_mean_token_accuracy": 0.7905156462533134,
+ "eval_num_tokens": 361815.0,
+ "eval_runtime": 57.6345,
+ "eval_samples_per_second": 24.222,
+ "eval_steps_per_second": 3.036,
+ "step": 140
+ },
+ {
+ "entropy": 0.7955982603132725,
+ "epoch": 0.3928790669122161,
+ "grad_norm": 0.5252285599708557,
+ "learning_rate": 0.00020923153562932872,
+ "loss": 0.7219765186309814,
+ "mean_token_accuracy": 0.792796092480421,
+ "num_tokens": 417429.0,
+ "step": 160
+ },
+ {
+ "epoch": 0.3928790669122161,
+ "eval_entropy": 0.7623447510174343,
+ "eval_loss": 0.7298487424850464,
+ "eval_mean_token_accuracy": 0.7945844326700483,
+ "eval_num_tokens": 417429.0,
+ "eval_runtime": 57.9123,
+ "eval_samples_per_second": 24.105,
+ "eval_steps_per_second": 3.022,
+ "step": 160
+ },
+ {
+ "entropy": 0.7903043553233147,
+ "epoch": 0.4419889502762431,
+ "grad_norm": 0.6948235034942627,
+ "learning_rate": 0.00020588431262670093,
+ "loss": 0.7225021839141845,
+ "mean_token_accuracy": 0.791443546116352,
+ "num_tokens": 467817.0,
+ "step": 180
+ },
+ {
+ "epoch": 0.4419889502762431,
+ "eval_entropy": 0.790012172971453,
+ "eval_loss": 0.720366895198822,
+ "eval_mean_token_accuracy": 0.7943330574035644,
+ "eval_num_tokens": 467817.0,
+ "eval_runtime": 57.4578,
+ "eval_samples_per_second": 24.296,
+ "eval_steps_per_second": 3.046,
+ "step": 180
+ },
+ {
+ "entropy": 0.7773297265172004,
+ "epoch": 0.4910988336402701,
+ "grad_norm": 0.6160358786582947,
+ "learning_rate": 0.00020181666283606295,
+ "loss": 0.7100065231323243,
+ "mean_token_accuracy": 0.7956355385482311,
+ "num_tokens": 513773.0,
+ "step": 200
+ },
+ {
+ "epoch": 0.4910988336402701,
+ "eval_entropy": 0.783927013874054,
+ "eval_loss": 0.7202737927436829,
+ "eval_mean_token_accuracy": 0.7964657868657793,
+ "eval_num_tokens": 513773.0,
+ "eval_runtime": 57.6663,
+ "eval_samples_per_second": 24.208,
+ "eval_steps_per_second": 3.035,
+ "step": 200
+ },
+ {
+ "entropy": 0.7645334511995315,
+ "epoch": 0.5402087170042971,
+ "grad_norm": 0.6620849967002869,
+ "learning_rate": 0.00019705837457374996,
+ "loss": 0.7037524223327637,
+ "mean_token_accuracy": 0.7996554173529148,
+ "num_tokens": 565606.0,
+ "step": 220
+ },
+ {
+ "epoch": 0.5402087170042971,
+ "eval_entropy": 0.7768624458994184,
+ "eval_loss": 0.7051481604576111,
+ "eval_mean_token_accuracy": 0.7991346962111336,
+ "eval_num_tokens": 565606.0,
+ "eval_runtime": 57.7442,
+ "eval_samples_per_second": 24.176,
+ "eval_steps_per_second": 3.031,
+ "step": 220
+ },
+ {
+ "entropy": 0.7731568738818169,
+ "epoch": 0.5893186003683242,
+ "grad_norm": 0.553688645362854,
+ "learning_rate": 0.00019164429385730688,
+ "loss": 0.7003726959228516,
+ "mean_token_accuracy": 0.7977605648338795,
+ "num_tokens": 616528.0,
+ "step": 240
+ },
+ {
+ "epoch": 0.5893186003683242,
+ "eval_entropy": 0.7570131056649344,
+ "eval_loss": 0.7042225003242493,
+ "eval_mean_token_accuracy": 0.8000479991095407,
+ "eval_num_tokens": 616528.0,
+ "eval_runtime": 57.8713,
+ "eval_samples_per_second": 24.122,
+ "eval_steps_per_second": 3.024,
+ "step": 240
+ },
+ {
+ "entropy": 0.7343776792287826,
+ "epoch": 0.6384284837323512,
+ "grad_norm": 0.5576229095458984,
+ "learning_rate": 0.00018561406922024723,
+ "loss": 0.6735261440277099,
+ "mean_token_accuracy": 0.8059290319681167,
+ "num_tokens": 669440.0,
+ "step": 260
+ },
+ {
+ "epoch": 0.6384284837323512,
+ "eval_entropy": 0.7231997234480722,
+ "eval_loss": 0.6926313638687134,
+ "eval_mean_token_accuracy": 0.8031776939119611,
+ "eval_num_tokens": 669440.0,
+ "eval_runtime": 57.1692,
+ "eval_samples_per_second": 24.419,
+ "eval_steps_per_second": 3.061,
+ "step": 260
+ },
+ {
+ "entropy": 0.7326226674020291,
+ "epoch": 0.6875383670963782,
+ "grad_norm": 0.5962154865264893,
+ "learning_rate": 0.00017901186135690458,
+ "loss": 0.6677989482879638,
+ "mean_token_accuracy": 0.8022022873163224,
+ "num_tokens": 720146.0,
+ "step": 280
+ },
+ {
+ "epoch": 0.6875383670963782,
+ "eval_entropy": 0.7265675282478332,
+ "eval_loss": 0.6867054104804993,
+ "eval_mean_token_accuracy": 0.8021906747136798,
+ "eval_num_tokens": 720146.0,
+ "eval_runtime": 57.8067,
+ "eval_samples_per_second": 24.149,
+ "eval_steps_per_second": 3.027,
+ "step": 280
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.204183549802496e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6f82990c2fbf5047472269cc86b324b35f1c571a
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/trainer_state.json
@@ -0,0 +1,349 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.7366482504604052,
+ "eval_steps": 20,
+ "global_step": 300,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ },
+ {
+ "entropy": 0.879152987897396,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 1.067221760749817,
+ "learning_rate": 0.00021473293557345806,
+ "loss": 0.807522201538086,
+ "mean_token_accuracy": 0.772921597212553,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.8607830974033901,
+ "eval_loss": 0.8036007881164551,
+ "eval_mean_token_accuracy": 0.7755460752759661,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 57.7576,
+ "eval_samples_per_second": 24.17,
+ "eval_steps_per_second": 3.03,
+ "step": 100
+ },
+ {
+ "entropy": 0.8589029781520366,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 0.7390042543411255,
+ "learning_rate": 0.0002136721067543722,
+ "loss": 0.7829637050628662,
+ "mean_token_accuracy": 0.779672809690237,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8330987681661334,
+ "eval_loss": 0.7742031216621399,
+ "eval_mean_token_accuracy": 0.785119104044778,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 57.3803,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 120
+ },
+ {
+ "entropy": 0.8094905942678452,
+ "epoch": 0.3437691835481891,
+ "grad_norm": 0.6055881381034851,
+ "learning_rate": 0.00021183381937537642,
+ "loss": 0.7396022796630859,
+ "mean_token_accuracy": 0.7867072127759457,
+ "num_tokens": 361815.0,
+ "step": 140
+ },
+ {
+ "epoch": 0.3437691835481891,
+ "eval_entropy": 0.7953300523757935,
+ "eval_loss": 0.7521983981132507,
+ "eval_mean_token_accuracy": 0.7905156462533134,
+ "eval_num_tokens": 361815.0,
+ "eval_runtime": 57.6345,
+ "eval_samples_per_second": 24.222,
+ "eval_steps_per_second": 3.036,
+ "step": 140
+ },
+ {
+ "entropy": 0.7955982603132725,
+ "epoch": 0.3928790669122161,
+ "grad_norm": 0.5252285599708557,
+ "learning_rate": 0.00020923153562932872,
+ "loss": 0.7219765186309814,
+ "mean_token_accuracy": 0.792796092480421,
+ "num_tokens": 417429.0,
+ "step": 160
+ },
+ {
+ "epoch": 0.3928790669122161,
+ "eval_entropy": 0.7623447510174343,
+ "eval_loss": 0.7298487424850464,
+ "eval_mean_token_accuracy": 0.7945844326700483,
+ "eval_num_tokens": 417429.0,
+ "eval_runtime": 57.9123,
+ "eval_samples_per_second": 24.105,
+ "eval_steps_per_second": 3.022,
+ "step": 160
+ },
+ {
+ "entropy": 0.7903043553233147,
+ "epoch": 0.4419889502762431,
+ "grad_norm": 0.6948235034942627,
+ "learning_rate": 0.00020588431262670093,
+ "loss": 0.7225021839141845,
+ "mean_token_accuracy": 0.791443546116352,
+ "num_tokens": 467817.0,
+ "step": 180
+ },
+ {
+ "epoch": 0.4419889502762431,
+ "eval_entropy": 0.790012172971453,
+ "eval_loss": 0.720366895198822,
+ "eval_mean_token_accuracy": 0.7943330574035644,
+ "eval_num_tokens": 467817.0,
+ "eval_runtime": 57.4578,
+ "eval_samples_per_second": 24.296,
+ "eval_steps_per_second": 3.046,
+ "step": 180
+ },
+ {
+ "entropy": 0.7773297265172004,
+ "epoch": 0.4910988336402701,
+ "grad_norm": 0.6160358786582947,
+ "learning_rate": 0.00020181666283606295,
+ "loss": 0.7100065231323243,
+ "mean_token_accuracy": 0.7956355385482311,
+ "num_tokens": 513773.0,
+ "step": 200
+ },
+ {
+ "epoch": 0.4910988336402701,
+ "eval_entropy": 0.783927013874054,
+ "eval_loss": 0.7202737927436829,
+ "eval_mean_token_accuracy": 0.7964657868657793,
+ "eval_num_tokens": 513773.0,
+ "eval_runtime": 57.6663,
+ "eval_samples_per_second": 24.208,
+ "eval_steps_per_second": 3.035,
+ "step": 200
+ },
+ {
+ "entropy": 0.7645334511995315,
+ "epoch": 0.5402087170042971,
+ "grad_norm": 0.6620849967002869,
+ "learning_rate": 0.00019705837457374996,
+ "loss": 0.7037524223327637,
+ "mean_token_accuracy": 0.7996554173529148,
+ "num_tokens": 565606.0,
+ "step": 220
+ },
+ {
+ "epoch": 0.5402087170042971,
+ "eval_entropy": 0.7768624458994184,
+ "eval_loss": 0.7051481604576111,
+ "eval_mean_token_accuracy": 0.7991346962111336,
+ "eval_num_tokens": 565606.0,
+ "eval_runtime": 57.7442,
+ "eval_samples_per_second": 24.176,
+ "eval_steps_per_second": 3.031,
+ "step": 220
+ },
+ {
+ "entropy": 0.7731568738818169,
+ "epoch": 0.5893186003683242,
+ "grad_norm": 0.553688645362854,
+ "learning_rate": 0.00019164429385730688,
+ "loss": 0.7003726959228516,
+ "mean_token_accuracy": 0.7977605648338795,
+ "num_tokens": 616528.0,
+ "step": 240
+ },
+ {
+ "epoch": 0.5893186003683242,
+ "eval_entropy": 0.7570131056649344,
+ "eval_loss": 0.7042225003242493,
+ "eval_mean_token_accuracy": 0.8000479991095407,
+ "eval_num_tokens": 616528.0,
+ "eval_runtime": 57.8713,
+ "eval_samples_per_second": 24.122,
+ "eval_steps_per_second": 3.024,
+ "step": 240
+ },
+ {
+ "entropy": 0.7343776792287826,
+ "epoch": 0.6384284837323512,
+ "grad_norm": 0.5576229095458984,
+ "learning_rate": 0.00018561406922024723,
+ "loss": 0.6735261440277099,
+ "mean_token_accuracy": 0.8059290319681167,
+ "num_tokens": 669440.0,
+ "step": 260
+ },
+ {
+ "epoch": 0.6384284837323512,
+ "eval_entropy": 0.7231997234480722,
+ "eval_loss": 0.6926313638687134,
+ "eval_mean_token_accuracy": 0.8031776939119611,
+ "eval_num_tokens": 669440.0,
+ "eval_runtime": 57.1692,
+ "eval_samples_per_second": 24.419,
+ "eval_steps_per_second": 3.061,
+ "step": 260
+ },
+ {
+ "entropy": 0.7326226674020291,
+ "epoch": 0.6875383670963782,
+ "grad_norm": 0.5962154865264893,
+ "learning_rate": 0.00017901186135690458,
+ "loss": 0.6677989482879638,
+ "mean_token_accuracy": 0.8022022873163224,
+ "num_tokens": 720146.0,
+ "step": 280
+ },
+ {
+ "epoch": 0.6875383670963782,
+ "eval_entropy": 0.7265675282478332,
+ "eval_loss": 0.6867054104804993,
+ "eval_mean_token_accuracy": 0.8021906747136798,
+ "eval_num_tokens": 720146.0,
+ "eval_runtime": 57.8067,
+ "eval_samples_per_second": 24.149,
+ "eval_steps_per_second": 3.027,
+ "step": 280
+ },
+ {
+ "entropy": 0.7409031428396702,
+ "epoch": 0.7366482504604052,
+ "grad_norm": 0.5716915726661682,
+ "learning_rate": 0.0001718860197237135,
+ "loss": 0.6692957878112793,
+ "mean_token_accuracy": 0.8036864519119262,
+ "num_tokens": 772624.0,
+ "step": 300
+ },
+ {
+ "epoch": 0.7366482504604052,
+ "eval_entropy": 0.7272419779641288,
+ "eval_loss": 0.6771560907363892,
+ "eval_mean_token_accuracy": 0.8062130059514727,
+ "eval_num_tokens": 772624.0,
+ "eval_runtime": 57.5942,
+ "eval_samples_per_second": 24.239,
+ "eval_steps_per_second": 3.039,
+ "step": 300
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.29431293498368e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..9471bb5534d783834146a64df9bdbf4ea1bae2a4
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/trainer_state.json
@@ -0,0 +1,370 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.7857581338244322,
+ "eval_steps": 20,
+ "global_step": 320,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ },
+ {
+ "entropy": 0.879152987897396,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 1.067221760749817,
+ "learning_rate": 0.00021473293557345806,
+ "loss": 0.807522201538086,
+ "mean_token_accuracy": 0.772921597212553,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.8607830974033901,
+ "eval_loss": 0.8036007881164551,
+ "eval_mean_token_accuracy": 0.7755460752759661,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 57.7576,
+ "eval_samples_per_second": 24.17,
+ "eval_steps_per_second": 3.03,
+ "step": 100
+ },
+ {
+ "entropy": 0.8589029781520366,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 0.7390042543411255,
+ "learning_rate": 0.0002136721067543722,
+ "loss": 0.7829637050628662,
+ "mean_token_accuracy": 0.779672809690237,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8330987681661334,
+ "eval_loss": 0.7742031216621399,
+ "eval_mean_token_accuracy": 0.785119104044778,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 57.3803,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 120
+ },
+ {
+ "entropy": 0.8094905942678452,
+ "epoch": 0.3437691835481891,
+ "grad_norm": 0.6055881381034851,
+ "learning_rate": 0.00021183381937537642,
+ "loss": 0.7396022796630859,
+ "mean_token_accuracy": 0.7867072127759457,
+ "num_tokens": 361815.0,
+ "step": 140
+ },
+ {
+ "epoch": 0.3437691835481891,
+ "eval_entropy": 0.7953300523757935,
+ "eval_loss": 0.7521983981132507,
+ "eval_mean_token_accuracy": 0.7905156462533134,
+ "eval_num_tokens": 361815.0,
+ "eval_runtime": 57.6345,
+ "eval_samples_per_second": 24.222,
+ "eval_steps_per_second": 3.036,
+ "step": 140
+ },
+ {
+ "entropy": 0.7955982603132725,
+ "epoch": 0.3928790669122161,
+ "grad_norm": 0.5252285599708557,
+ "learning_rate": 0.00020923153562932872,
+ "loss": 0.7219765186309814,
+ "mean_token_accuracy": 0.792796092480421,
+ "num_tokens": 417429.0,
+ "step": 160
+ },
+ {
+ "epoch": 0.3928790669122161,
+ "eval_entropy": 0.7623447510174343,
+ "eval_loss": 0.7298487424850464,
+ "eval_mean_token_accuracy": 0.7945844326700483,
+ "eval_num_tokens": 417429.0,
+ "eval_runtime": 57.9123,
+ "eval_samples_per_second": 24.105,
+ "eval_steps_per_second": 3.022,
+ "step": 160
+ },
+ {
+ "entropy": 0.7903043553233147,
+ "epoch": 0.4419889502762431,
+ "grad_norm": 0.6948235034942627,
+ "learning_rate": 0.00020588431262670093,
+ "loss": 0.7225021839141845,
+ "mean_token_accuracy": 0.791443546116352,
+ "num_tokens": 467817.0,
+ "step": 180
+ },
+ {
+ "epoch": 0.4419889502762431,
+ "eval_entropy": 0.790012172971453,
+ "eval_loss": 0.720366895198822,
+ "eval_mean_token_accuracy": 0.7943330574035644,
+ "eval_num_tokens": 467817.0,
+ "eval_runtime": 57.4578,
+ "eval_samples_per_second": 24.296,
+ "eval_steps_per_second": 3.046,
+ "step": 180
+ },
+ {
+ "entropy": 0.7773297265172004,
+ "epoch": 0.4910988336402701,
+ "grad_norm": 0.6160358786582947,
+ "learning_rate": 0.00020181666283606295,
+ "loss": 0.7100065231323243,
+ "mean_token_accuracy": 0.7956355385482311,
+ "num_tokens": 513773.0,
+ "step": 200
+ },
+ {
+ "epoch": 0.4910988336402701,
+ "eval_entropy": 0.783927013874054,
+ "eval_loss": 0.7202737927436829,
+ "eval_mean_token_accuracy": 0.7964657868657793,
+ "eval_num_tokens": 513773.0,
+ "eval_runtime": 57.6663,
+ "eval_samples_per_second": 24.208,
+ "eval_steps_per_second": 3.035,
+ "step": 200
+ },
+ {
+ "entropy": 0.7645334511995315,
+ "epoch": 0.5402087170042971,
+ "grad_norm": 0.6620849967002869,
+ "learning_rate": 0.00019705837457374996,
+ "loss": 0.7037524223327637,
+ "mean_token_accuracy": 0.7996554173529148,
+ "num_tokens": 565606.0,
+ "step": 220
+ },
+ {
+ "epoch": 0.5402087170042971,
+ "eval_entropy": 0.7768624458994184,
+ "eval_loss": 0.7051481604576111,
+ "eval_mean_token_accuracy": 0.7991346962111336,
+ "eval_num_tokens": 565606.0,
+ "eval_runtime": 57.7442,
+ "eval_samples_per_second": 24.176,
+ "eval_steps_per_second": 3.031,
+ "step": 220
+ },
+ {
+ "entropy": 0.7731568738818169,
+ "epoch": 0.5893186003683242,
+ "grad_norm": 0.553688645362854,
+ "learning_rate": 0.00019164429385730688,
+ "loss": 0.7003726959228516,
+ "mean_token_accuracy": 0.7977605648338795,
+ "num_tokens": 616528.0,
+ "step": 240
+ },
+ {
+ "epoch": 0.5893186003683242,
+ "eval_entropy": 0.7570131056649344,
+ "eval_loss": 0.7042225003242493,
+ "eval_mean_token_accuracy": 0.8000479991095407,
+ "eval_num_tokens": 616528.0,
+ "eval_runtime": 57.8713,
+ "eval_samples_per_second": 24.122,
+ "eval_steps_per_second": 3.024,
+ "step": 240
+ },
+ {
+ "entropy": 0.7343776792287826,
+ "epoch": 0.6384284837323512,
+ "grad_norm": 0.5576229095458984,
+ "learning_rate": 0.00018561406922024723,
+ "loss": 0.6735261440277099,
+ "mean_token_accuracy": 0.8059290319681167,
+ "num_tokens": 669440.0,
+ "step": 260
+ },
+ {
+ "epoch": 0.6384284837323512,
+ "eval_entropy": 0.7231997234480722,
+ "eval_loss": 0.6926313638687134,
+ "eval_mean_token_accuracy": 0.8031776939119611,
+ "eval_num_tokens": 669440.0,
+ "eval_runtime": 57.1692,
+ "eval_samples_per_second": 24.419,
+ "eval_steps_per_second": 3.061,
+ "step": 260
+ },
+ {
+ "entropy": 0.7326226674020291,
+ "epoch": 0.6875383670963782,
+ "grad_norm": 0.5962154865264893,
+ "learning_rate": 0.00017901186135690458,
+ "loss": 0.6677989482879638,
+ "mean_token_accuracy": 0.8022022873163224,
+ "num_tokens": 720146.0,
+ "step": 280
+ },
+ {
+ "epoch": 0.6875383670963782,
+ "eval_entropy": 0.7265675282478332,
+ "eval_loss": 0.6867054104804993,
+ "eval_mean_token_accuracy": 0.8021906747136798,
+ "eval_num_tokens": 720146.0,
+ "eval_runtime": 57.8067,
+ "eval_samples_per_second": 24.149,
+ "eval_steps_per_second": 3.027,
+ "step": 280
+ },
+ {
+ "entropy": 0.7409031428396702,
+ "epoch": 0.7366482504604052,
+ "grad_norm": 0.5716915726661682,
+ "learning_rate": 0.0001718860197237135,
+ "loss": 0.6692957878112793,
+ "mean_token_accuracy": 0.8036864519119262,
+ "num_tokens": 772624.0,
+ "step": 300
+ },
+ {
+ "epoch": 0.7366482504604052,
+ "eval_entropy": 0.7272419779641288,
+ "eval_loss": 0.6771560907363892,
+ "eval_mean_token_accuracy": 0.8062130059514727,
+ "eval_num_tokens": 772624.0,
+ "eval_runtime": 57.5942,
+ "eval_samples_per_second": 24.239,
+ "eval_steps_per_second": 3.039,
+ "step": 300
+ },
+ {
+ "entropy": 0.7243741095066071,
+ "epoch": 0.7857581338244322,
+ "grad_norm": 0.5360459089279175,
+ "learning_rate": 0.00016428872846524118,
+ "loss": 0.6628021240234375,
+ "mean_token_accuracy": 0.8050650551915168,
+ "num_tokens": 822969.0,
+ "step": 320
+ },
+ {
+ "epoch": 0.7857581338244322,
+ "eval_entropy": 0.7325100105149406,
+ "eval_loss": 0.6667791604995728,
+ "eval_mean_token_accuracy": 0.8082476602281843,
+ "eval_num_tokens": 822969.0,
+ "eval_runtime": 57.3807,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 320
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.379271938897664e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..8f1f23e179a1f64641a8fc0eb6653e18a4730eab
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/trainer_state.json
@@ -0,0 +1,391 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.8348680171884592,
+ "eval_steps": 20,
+ "global_step": 340,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ },
+ {
+ "entropy": 0.879152987897396,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 1.067221760749817,
+ "learning_rate": 0.00021473293557345806,
+ "loss": 0.807522201538086,
+ "mean_token_accuracy": 0.772921597212553,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.8607830974033901,
+ "eval_loss": 0.8036007881164551,
+ "eval_mean_token_accuracy": 0.7755460752759661,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 57.7576,
+ "eval_samples_per_second": 24.17,
+ "eval_steps_per_second": 3.03,
+ "step": 100
+ },
+ {
+ "entropy": 0.8589029781520366,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 0.7390042543411255,
+ "learning_rate": 0.0002136721067543722,
+ "loss": 0.7829637050628662,
+ "mean_token_accuracy": 0.779672809690237,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8330987681661334,
+ "eval_loss": 0.7742031216621399,
+ "eval_mean_token_accuracy": 0.785119104044778,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 57.3803,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 120
+ },
+ {
+ "entropy": 0.8094905942678452,
+ "epoch": 0.3437691835481891,
+ "grad_norm": 0.6055881381034851,
+ "learning_rate": 0.00021183381937537642,
+ "loss": 0.7396022796630859,
+ "mean_token_accuracy": 0.7867072127759457,
+ "num_tokens": 361815.0,
+ "step": 140
+ },
+ {
+ "epoch": 0.3437691835481891,
+ "eval_entropy": 0.7953300523757935,
+ "eval_loss": 0.7521983981132507,
+ "eval_mean_token_accuracy": 0.7905156462533134,
+ "eval_num_tokens": 361815.0,
+ "eval_runtime": 57.6345,
+ "eval_samples_per_second": 24.222,
+ "eval_steps_per_second": 3.036,
+ "step": 140
+ },
+ {
+ "entropy": 0.7955982603132725,
+ "epoch": 0.3928790669122161,
+ "grad_norm": 0.5252285599708557,
+ "learning_rate": 0.00020923153562932872,
+ "loss": 0.7219765186309814,
+ "mean_token_accuracy": 0.792796092480421,
+ "num_tokens": 417429.0,
+ "step": 160
+ },
+ {
+ "epoch": 0.3928790669122161,
+ "eval_entropy": 0.7623447510174343,
+ "eval_loss": 0.7298487424850464,
+ "eval_mean_token_accuracy": 0.7945844326700483,
+ "eval_num_tokens": 417429.0,
+ "eval_runtime": 57.9123,
+ "eval_samples_per_second": 24.105,
+ "eval_steps_per_second": 3.022,
+ "step": 160
+ },
+ {
+ "entropy": 0.7903043553233147,
+ "epoch": 0.4419889502762431,
+ "grad_norm": 0.6948235034942627,
+ "learning_rate": 0.00020588431262670093,
+ "loss": 0.7225021839141845,
+ "mean_token_accuracy": 0.791443546116352,
+ "num_tokens": 467817.0,
+ "step": 180
+ },
+ {
+ "epoch": 0.4419889502762431,
+ "eval_entropy": 0.790012172971453,
+ "eval_loss": 0.720366895198822,
+ "eval_mean_token_accuracy": 0.7943330574035644,
+ "eval_num_tokens": 467817.0,
+ "eval_runtime": 57.4578,
+ "eval_samples_per_second": 24.296,
+ "eval_steps_per_second": 3.046,
+ "step": 180
+ },
+ {
+ "entropy": 0.7773297265172004,
+ "epoch": 0.4910988336402701,
+ "grad_norm": 0.6160358786582947,
+ "learning_rate": 0.00020181666283606295,
+ "loss": 0.7100065231323243,
+ "mean_token_accuracy": 0.7956355385482311,
+ "num_tokens": 513773.0,
+ "step": 200
+ },
+ {
+ "epoch": 0.4910988336402701,
+ "eval_entropy": 0.783927013874054,
+ "eval_loss": 0.7202737927436829,
+ "eval_mean_token_accuracy": 0.7964657868657793,
+ "eval_num_tokens": 513773.0,
+ "eval_runtime": 57.6663,
+ "eval_samples_per_second": 24.208,
+ "eval_steps_per_second": 3.035,
+ "step": 200
+ },
+ {
+ "entropy": 0.7645334511995315,
+ "epoch": 0.5402087170042971,
+ "grad_norm": 0.6620849967002869,
+ "learning_rate": 0.00019705837457374996,
+ "loss": 0.7037524223327637,
+ "mean_token_accuracy": 0.7996554173529148,
+ "num_tokens": 565606.0,
+ "step": 220
+ },
+ {
+ "epoch": 0.5402087170042971,
+ "eval_entropy": 0.7768624458994184,
+ "eval_loss": 0.7051481604576111,
+ "eval_mean_token_accuracy": 0.7991346962111336,
+ "eval_num_tokens": 565606.0,
+ "eval_runtime": 57.7442,
+ "eval_samples_per_second": 24.176,
+ "eval_steps_per_second": 3.031,
+ "step": 220
+ },
+ {
+ "entropy": 0.7731568738818169,
+ "epoch": 0.5893186003683242,
+ "grad_norm": 0.553688645362854,
+ "learning_rate": 0.00019164429385730688,
+ "loss": 0.7003726959228516,
+ "mean_token_accuracy": 0.7977605648338795,
+ "num_tokens": 616528.0,
+ "step": 240
+ },
+ {
+ "epoch": 0.5893186003683242,
+ "eval_entropy": 0.7570131056649344,
+ "eval_loss": 0.7042225003242493,
+ "eval_mean_token_accuracy": 0.8000479991095407,
+ "eval_num_tokens": 616528.0,
+ "eval_runtime": 57.8713,
+ "eval_samples_per_second": 24.122,
+ "eval_steps_per_second": 3.024,
+ "step": 240
+ },
+ {
+ "entropy": 0.7343776792287826,
+ "epoch": 0.6384284837323512,
+ "grad_norm": 0.5576229095458984,
+ "learning_rate": 0.00018561406922024723,
+ "loss": 0.6735261440277099,
+ "mean_token_accuracy": 0.8059290319681167,
+ "num_tokens": 669440.0,
+ "step": 260
+ },
+ {
+ "epoch": 0.6384284837323512,
+ "eval_entropy": 0.7231997234480722,
+ "eval_loss": 0.6926313638687134,
+ "eval_mean_token_accuracy": 0.8031776939119611,
+ "eval_num_tokens": 669440.0,
+ "eval_runtime": 57.1692,
+ "eval_samples_per_second": 24.419,
+ "eval_steps_per_second": 3.061,
+ "step": 260
+ },
+ {
+ "entropy": 0.7326226674020291,
+ "epoch": 0.6875383670963782,
+ "grad_norm": 0.5962154865264893,
+ "learning_rate": 0.00017901186135690458,
+ "loss": 0.6677989482879638,
+ "mean_token_accuracy": 0.8022022873163224,
+ "num_tokens": 720146.0,
+ "step": 280
+ },
+ {
+ "epoch": 0.6875383670963782,
+ "eval_entropy": 0.7265675282478332,
+ "eval_loss": 0.6867054104804993,
+ "eval_mean_token_accuracy": 0.8021906747136798,
+ "eval_num_tokens": 720146.0,
+ "eval_runtime": 57.8067,
+ "eval_samples_per_second": 24.149,
+ "eval_steps_per_second": 3.027,
+ "step": 280
+ },
+ {
+ "entropy": 0.7409031428396702,
+ "epoch": 0.7366482504604052,
+ "grad_norm": 0.5716915726661682,
+ "learning_rate": 0.0001718860197237135,
+ "loss": 0.6692957878112793,
+ "mean_token_accuracy": 0.8036864519119262,
+ "num_tokens": 772624.0,
+ "step": 300
+ },
+ {
+ "epoch": 0.7366482504604052,
+ "eval_entropy": 0.7272419779641288,
+ "eval_loss": 0.6771560907363892,
+ "eval_mean_token_accuracy": 0.8062130059514727,
+ "eval_num_tokens": 772624.0,
+ "eval_runtime": 57.5942,
+ "eval_samples_per_second": 24.239,
+ "eval_steps_per_second": 3.039,
+ "step": 300
+ },
+ {
+ "entropy": 0.7243741095066071,
+ "epoch": 0.7857581338244322,
+ "grad_norm": 0.5360459089279175,
+ "learning_rate": 0.00016428872846524118,
+ "loss": 0.6628021240234375,
+ "mean_token_accuracy": 0.8050650551915168,
+ "num_tokens": 822969.0,
+ "step": 320
+ },
+ {
+ "epoch": 0.7857581338244322,
+ "eval_entropy": 0.7325100105149406,
+ "eval_loss": 0.6667791604995728,
+ "eval_mean_token_accuracy": 0.8082476602281843,
+ "eval_num_tokens": 822969.0,
+ "eval_runtime": 57.3807,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 320
+ },
+ {
+ "entropy": 0.7213595665991306,
+ "epoch": 0.8348680171884592,
+ "grad_norm": 0.48658841848373413,
+ "learning_rate": 0.00015627562425793356,
+ "loss": 0.6494232654571533,
+ "mean_token_accuracy": 0.811632139235735,
+ "num_tokens": 873459.0,
+ "step": 340
+ },
+ {
+ "epoch": 0.8348680171884592,
+ "eval_entropy": 0.7198746153286525,
+ "eval_loss": 0.6634471416473389,
+ "eval_mean_token_accuracy": 0.8094694914136614,
+ "eval_num_tokens": 873459.0,
+ "eval_runtime": 57.6988,
+ "eval_samples_per_second": 24.195,
+ "eval_steps_per_second": 3.033,
+ "step": 340
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.466169397618944e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..492d50334065285f77598321e19501a0f79f767e
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/trainer_state.json
@@ -0,0 +1,412 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.8839779005524862,
+ "eval_steps": 20,
+ "global_step": 360,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ },
+ {
+ "entropy": 0.879152987897396,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 1.067221760749817,
+ "learning_rate": 0.00021473293557345806,
+ "loss": 0.807522201538086,
+ "mean_token_accuracy": 0.772921597212553,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.8607830974033901,
+ "eval_loss": 0.8036007881164551,
+ "eval_mean_token_accuracy": 0.7755460752759661,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 57.7576,
+ "eval_samples_per_second": 24.17,
+ "eval_steps_per_second": 3.03,
+ "step": 100
+ },
+ {
+ "entropy": 0.8589029781520366,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 0.7390042543411255,
+ "learning_rate": 0.0002136721067543722,
+ "loss": 0.7829637050628662,
+ "mean_token_accuracy": 0.779672809690237,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8330987681661334,
+ "eval_loss": 0.7742031216621399,
+ "eval_mean_token_accuracy": 0.785119104044778,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 57.3803,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 120
+ },
+ {
+ "entropy": 0.8094905942678452,
+ "epoch": 0.3437691835481891,
+ "grad_norm": 0.6055881381034851,
+ "learning_rate": 0.00021183381937537642,
+ "loss": 0.7396022796630859,
+ "mean_token_accuracy": 0.7867072127759457,
+ "num_tokens": 361815.0,
+ "step": 140
+ },
+ {
+ "epoch": 0.3437691835481891,
+ "eval_entropy": 0.7953300523757935,
+ "eval_loss": 0.7521983981132507,
+ "eval_mean_token_accuracy": 0.7905156462533134,
+ "eval_num_tokens": 361815.0,
+ "eval_runtime": 57.6345,
+ "eval_samples_per_second": 24.222,
+ "eval_steps_per_second": 3.036,
+ "step": 140
+ },
+ {
+ "entropy": 0.7955982603132725,
+ "epoch": 0.3928790669122161,
+ "grad_norm": 0.5252285599708557,
+ "learning_rate": 0.00020923153562932872,
+ "loss": 0.7219765186309814,
+ "mean_token_accuracy": 0.792796092480421,
+ "num_tokens": 417429.0,
+ "step": 160
+ },
+ {
+ "epoch": 0.3928790669122161,
+ "eval_entropy": 0.7623447510174343,
+ "eval_loss": 0.7298487424850464,
+ "eval_mean_token_accuracy": 0.7945844326700483,
+ "eval_num_tokens": 417429.0,
+ "eval_runtime": 57.9123,
+ "eval_samples_per_second": 24.105,
+ "eval_steps_per_second": 3.022,
+ "step": 160
+ },
+ {
+ "entropy": 0.7903043553233147,
+ "epoch": 0.4419889502762431,
+ "grad_norm": 0.6948235034942627,
+ "learning_rate": 0.00020588431262670093,
+ "loss": 0.7225021839141845,
+ "mean_token_accuracy": 0.791443546116352,
+ "num_tokens": 467817.0,
+ "step": 180
+ },
+ {
+ "epoch": 0.4419889502762431,
+ "eval_entropy": 0.790012172971453,
+ "eval_loss": 0.720366895198822,
+ "eval_mean_token_accuracy": 0.7943330574035644,
+ "eval_num_tokens": 467817.0,
+ "eval_runtime": 57.4578,
+ "eval_samples_per_second": 24.296,
+ "eval_steps_per_second": 3.046,
+ "step": 180
+ },
+ {
+ "entropy": 0.7773297265172004,
+ "epoch": 0.4910988336402701,
+ "grad_norm": 0.6160358786582947,
+ "learning_rate": 0.00020181666283606295,
+ "loss": 0.7100065231323243,
+ "mean_token_accuracy": 0.7956355385482311,
+ "num_tokens": 513773.0,
+ "step": 200
+ },
+ {
+ "epoch": 0.4910988336402701,
+ "eval_entropy": 0.783927013874054,
+ "eval_loss": 0.7202737927436829,
+ "eval_mean_token_accuracy": 0.7964657868657793,
+ "eval_num_tokens": 513773.0,
+ "eval_runtime": 57.6663,
+ "eval_samples_per_second": 24.208,
+ "eval_steps_per_second": 3.035,
+ "step": 200
+ },
+ {
+ "entropy": 0.7645334511995315,
+ "epoch": 0.5402087170042971,
+ "grad_norm": 0.6620849967002869,
+ "learning_rate": 0.00019705837457374996,
+ "loss": 0.7037524223327637,
+ "mean_token_accuracy": 0.7996554173529148,
+ "num_tokens": 565606.0,
+ "step": 220
+ },
+ {
+ "epoch": 0.5402087170042971,
+ "eval_entropy": 0.7768624458994184,
+ "eval_loss": 0.7051481604576111,
+ "eval_mean_token_accuracy": 0.7991346962111336,
+ "eval_num_tokens": 565606.0,
+ "eval_runtime": 57.7442,
+ "eval_samples_per_second": 24.176,
+ "eval_steps_per_second": 3.031,
+ "step": 220
+ },
+ {
+ "entropy": 0.7731568738818169,
+ "epoch": 0.5893186003683242,
+ "grad_norm": 0.553688645362854,
+ "learning_rate": 0.00019164429385730688,
+ "loss": 0.7003726959228516,
+ "mean_token_accuracy": 0.7977605648338795,
+ "num_tokens": 616528.0,
+ "step": 240
+ },
+ {
+ "epoch": 0.5893186003683242,
+ "eval_entropy": 0.7570131056649344,
+ "eval_loss": 0.7042225003242493,
+ "eval_mean_token_accuracy": 0.8000479991095407,
+ "eval_num_tokens": 616528.0,
+ "eval_runtime": 57.8713,
+ "eval_samples_per_second": 24.122,
+ "eval_steps_per_second": 3.024,
+ "step": 240
+ },
+ {
+ "entropy": 0.7343776792287826,
+ "epoch": 0.6384284837323512,
+ "grad_norm": 0.5576229095458984,
+ "learning_rate": 0.00018561406922024723,
+ "loss": 0.6735261440277099,
+ "mean_token_accuracy": 0.8059290319681167,
+ "num_tokens": 669440.0,
+ "step": 260
+ },
+ {
+ "epoch": 0.6384284837323512,
+ "eval_entropy": 0.7231997234480722,
+ "eval_loss": 0.6926313638687134,
+ "eval_mean_token_accuracy": 0.8031776939119611,
+ "eval_num_tokens": 669440.0,
+ "eval_runtime": 57.1692,
+ "eval_samples_per_second": 24.419,
+ "eval_steps_per_second": 3.061,
+ "step": 260
+ },
+ {
+ "entropy": 0.7326226674020291,
+ "epoch": 0.6875383670963782,
+ "grad_norm": 0.5962154865264893,
+ "learning_rate": 0.00017901186135690458,
+ "loss": 0.6677989482879638,
+ "mean_token_accuracy": 0.8022022873163224,
+ "num_tokens": 720146.0,
+ "step": 280
+ },
+ {
+ "epoch": 0.6875383670963782,
+ "eval_entropy": 0.7265675282478332,
+ "eval_loss": 0.6867054104804993,
+ "eval_mean_token_accuracy": 0.8021906747136798,
+ "eval_num_tokens": 720146.0,
+ "eval_runtime": 57.8067,
+ "eval_samples_per_second": 24.149,
+ "eval_steps_per_second": 3.027,
+ "step": 280
+ },
+ {
+ "entropy": 0.7409031428396702,
+ "epoch": 0.7366482504604052,
+ "grad_norm": 0.5716915726661682,
+ "learning_rate": 0.0001718860197237135,
+ "loss": 0.6692957878112793,
+ "mean_token_accuracy": 0.8036864519119262,
+ "num_tokens": 772624.0,
+ "step": 300
+ },
+ {
+ "epoch": 0.7366482504604052,
+ "eval_entropy": 0.7272419779641288,
+ "eval_loss": 0.6771560907363892,
+ "eval_mean_token_accuracy": 0.8062130059514727,
+ "eval_num_tokens": 772624.0,
+ "eval_runtime": 57.5942,
+ "eval_samples_per_second": 24.239,
+ "eval_steps_per_second": 3.039,
+ "step": 300
+ },
+ {
+ "entropy": 0.7243741095066071,
+ "epoch": 0.7857581338244322,
+ "grad_norm": 0.5360459089279175,
+ "learning_rate": 0.00016428872846524118,
+ "loss": 0.6628021240234375,
+ "mean_token_accuracy": 0.8050650551915168,
+ "num_tokens": 822969.0,
+ "step": 320
+ },
+ {
+ "epoch": 0.7857581338244322,
+ "eval_entropy": 0.7325100105149406,
+ "eval_loss": 0.6667791604995728,
+ "eval_mean_token_accuracy": 0.8082476602281843,
+ "eval_num_tokens": 822969.0,
+ "eval_runtime": 57.3807,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 320
+ },
+ {
+ "entropy": 0.7213595665991306,
+ "epoch": 0.8348680171884592,
+ "grad_norm": 0.48658841848373413,
+ "learning_rate": 0.00015627562425793356,
+ "loss": 0.6494232654571533,
+ "mean_token_accuracy": 0.811632139235735,
+ "num_tokens": 873459.0,
+ "step": 340
+ },
+ {
+ "epoch": 0.8348680171884592,
+ "eval_entropy": 0.7198746153286525,
+ "eval_loss": 0.6634471416473389,
+ "eval_mean_token_accuracy": 0.8094694914136614,
+ "eval_num_tokens": 873459.0,
+ "eval_runtime": 57.6988,
+ "eval_samples_per_second": 24.195,
+ "eval_steps_per_second": 3.033,
+ "step": 340
+ },
+ {
+ "entropy": 0.7133754774928093,
+ "epoch": 0.8839779005524862,
+ "grad_norm": 0.685793399810791,
+ "learning_rate": 0.00014790538887019238,
+ "loss": 0.642530870437622,
+ "mean_token_accuracy": 0.8122274152934551,
+ "num_tokens": 925669.0,
+ "step": 360
+ },
+ {
+ "epoch": 0.8839779005524862,
+ "eval_entropy": 0.6683637884684971,
+ "eval_loss": 0.6579923033714294,
+ "eval_mean_token_accuracy": 0.811241352558136,
+ "eval_num_tokens": 925669.0,
+ "eval_runtime": 57.8503,
+ "eval_samples_per_second": 24.131,
+ "eval_steps_per_second": 3.025,
+ "step": 360
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.553852053224192e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..df52cc65a6a7b4bd7e732b12cd9289b0ebfdeae5
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/trainer_state.json
@@ -0,0 +1,433 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.9330877839165131,
+ "eval_steps": 20,
+ "global_step": 380,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ },
+ {
+ "entropy": 0.879152987897396,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 1.067221760749817,
+ "learning_rate": 0.00021473293557345806,
+ "loss": 0.807522201538086,
+ "mean_token_accuracy": 0.772921597212553,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.8607830974033901,
+ "eval_loss": 0.8036007881164551,
+ "eval_mean_token_accuracy": 0.7755460752759661,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 57.7576,
+ "eval_samples_per_second": 24.17,
+ "eval_steps_per_second": 3.03,
+ "step": 100
+ },
+ {
+ "entropy": 0.8589029781520366,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 0.7390042543411255,
+ "learning_rate": 0.0002136721067543722,
+ "loss": 0.7829637050628662,
+ "mean_token_accuracy": 0.779672809690237,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8330987681661334,
+ "eval_loss": 0.7742031216621399,
+ "eval_mean_token_accuracy": 0.785119104044778,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 57.3803,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 120
+ },
+ {
+ "entropy": 0.8094905942678452,
+ "epoch": 0.3437691835481891,
+ "grad_norm": 0.6055881381034851,
+ "learning_rate": 0.00021183381937537642,
+ "loss": 0.7396022796630859,
+ "mean_token_accuracy": 0.7867072127759457,
+ "num_tokens": 361815.0,
+ "step": 140
+ },
+ {
+ "epoch": 0.3437691835481891,
+ "eval_entropy": 0.7953300523757935,
+ "eval_loss": 0.7521983981132507,
+ "eval_mean_token_accuracy": 0.7905156462533134,
+ "eval_num_tokens": 361815.0,
+ "eval_runtime": 57.6345,
+ "eval_samples_per_second": 24.222,
+ "eval_steps_per_second": 3.036,
+ "step": 140
+ },
+ {
+ "entropy": 0.7955982603132725,
+ "epoch": 0.3928790669122161,
+ "grad_norm": 0.5252285599708557,
+ "learning_rate": 0.00020923153562932872,
+ "loss": 0.7219765186309814,
+ "mean_token_accuracy": 0.792796092480421,
+ "num_tokens": 417429.0,
+ "step": 160
+ },
+ {
+ "epoch": 0.3928790669122161,
+ "eval_entropy": 0.7623447510174343,
+ "eval_loss": 0.7298487424850464,
+ "eval_mean_token_accuracy": 0.7945844326700483,
+ "eval_num_tokens": 417429.0,
+ "eval_runtime": 57.9123,
+ "eval_samples_per_second": 24.105,
+ "eval_steps_per_second": 3.022,
+ "step": 160
+ },
+ {
+ "entropy": 0.7903043553233147,
+ "epoch": 0.4419889502762431,
+ "grad_norm": 0.6948235034942627,
+ "learning_rate": 0.00020588431262670093,
+ "loss": 0.7225021839141845,
+ "mean_token_accuracy": 0.791443546116352,
+ "num_tokens": 467817.0,
+ "step": 180
+ },
+ {
+ "epoch": 0.4419889502762431,
+ "eval_entropy": 0.790012172971453,
+ "eval_loss": 0.720366895198822,
+ "eval_mean_token_accuracy": 0.7943330574035644,
+ "eval_num_tokens": 467817.0,
+ "eval_runtime": 57.4578,
+ "eval_samples_per_second": 24.296,
+ "eval_steps_per_second": 3.046,
+ "step": 180
+ },
+ {
+ "entropy": 0.7773297265172004,
+ "epoch": 0.4910988336402701,
+ "grad_norm": 0.6160358786582947,
+ "learning_rate": 0.00020181666283606295,
+ "loss": 0.7100065231323243,
+ "mean_token_accuracy": 0.7956355385482311,
+ "num_tokens": 513773.0,
+ "step": 200
+ },
+ {
+ "epoch": 0.4910988336402701,
+ "eval_entropy": 0.783927013874054,
+ "eval_loss": 0.7202737927436829,
+ "eval_mean_token_accuracy": 0.7964657868657793,
+ "eval_num_tokens": 513773.0,
+ "eval_runtime": 57.6663,
+ "eval_samples_per_second": 24.208,
+ "eval_steps_per_second": 3.035,
+ "step": 200
+ },
+ {
+ "entropy": 0.7645334511995315,
+ "epoch": 0.5402087170042971,
+ "grad_norm": 0.6620849967002869,
+ "learning_rate": 0.00019705837457374996,
+ "loss": 0.7037524223327637,
+ "mean_token_accuracy": 0.7996554173529148,
+ "num_tokens": 565606.0,
+ "step": 220
+ },
+ {
+ "epoch": 0.5402087170042971,
+ "eval_entropy": 0.7768624458994184,
+ "eval_loss": 0.7051481604576111,
+ "eval_mean_token_accuracy": 0.7991346962111336,
+ "eval_num_tokens": 565606.0,
+ "eval_runtime": 57.7442,
+ "eval_samples_per_second": 24.176,
+ "eval_steps_per_second": 3.031,
+ "step": 220
+ },
+ {
+ "entropy": 0.7731568738818169,
+ "epoch": 0.5893186003683242,
+ "grad_norm": 0.553688645362854,
+ "learning_rate": 0.00019164429385730688,
+ "loss": 0.7003726959228516,
+ "mean_token_accuracy": 0.7977605648338795,
+ "num_tokens": 616528.0,
+ "step": 240
+ },
+ {
+ "epoch": 0.5893186003683242,
+ "eval_entropy": 0.7570131056649344,
+ "eval_loss": 0.7042225003242493,
+ "eval_mean_token_accuracy": 0.8000479991095407,
+ "eval_num_tokens": 616528.0,
+ "eval_runtime": 57.8713,
+ "eval_samples_per_second": 24.122,
+ "eval_steps_per_second": 3.024,
+ "step": 240
+ },
+ {
+ "entropy": 0.7343776792287826,
+ "epoch": 0.6384284837323512,
+ "grad_norm": 0.5576229095458984,
+ "learning_rate": 0.00018561406922024723,
+ "loss": 0.6735261440277099,
+ "mean_token_accuracy": 0.8059290319681167,
+ "num_tokens": 669440.0,
+ "step": 260
+ },
+ {
+ "epoch": 0.6384284837323512,
+ "eval_entropy": 0.7231997234480722,
+ "eval_loss": 0.6926313638687134,
+ "eval_mean_token_accuracy": 0.8031776939119611,
+ "eval_num_tokens": 669440.0,
+ "eval_runtime": 57.1692,
+ "eval_samples_per_second": 24.419,
+ "eval_steps_per_second": 3.061,
+ "step": 260
+ },
+ {
+ "entropy": 0.7326226674020291,
+ "epoch": 0.6875383670963782,
+ "grad_norm": 0.5962154865264893,
+ "learning_rate": 0.00017901186135690458,
+ "loss": 0.6677989482879638,
+ "mean_token_accuracy": 0.8022022873163224,
+ "num_tokens": 720146.0,
+ "step": 280
+ },
+ {
+ "epoch": 0.6875383670963782,
+ "eval_entropy": 0.7265675282478332,
+ "eval_loss": 0.6867054104804993,
+ "eval_mean_token_accuracy": 0.8021906747136798,
+ "eval_num_tokens": 720146.0,
+ "eval_runtime": 57.8067,
+ "eval_samples_per_second": 24.149,
+ "eval_steps_per_second": 3.027,
+ "step": 280
+ },
+ {
+ "entropy": 0.7409031428396702,
+ "epoch": 0.7366482504604052,
+ "grad_norm": 0.5716915726661682,
+ "learning_rate": 0.0001718860197237135,
+ "loss": 0.6692957878112793,
+ "mean_token_accuracy": 0.8036864519119262,
+ "num_tokens": 772624.0,
+ "step": 300
+ },
+ {
+ "epoch": 0.7366482504604052,
+ "eval_entropy": 0.7272419779641288,
+ "eval_loss": 0.6771560907363892,
+ "eval_mean_token_accuracy": 0.8062130059514727,
+ "eval_num_tokens": 772624.0,
+ "eval_runtime": 57.5942,
+ "eval_samples_per_second": 24.239,
+ "eval_steps_per_second": 3.039,
+ "step": 300
+ },
+ {
+ "entropy": 0.7243741095066071,
+ "epoch": 0.7857581338244322,
+ "grad_norm": 0.5360459089279175,
+ "learning_rate": 0.00016428872846524118,
+ "loss": 0.6628021240234375,
+ "mean_token_accuracy": 0.8050650551915168,
+ "num_tokens": 822969.0,
+ "step": 320
+ },
+ {
+ "epoch": 0.7857581338244322,
+ "eval_entropy": 0.7325100105149406,
+ "eval_loss": 0.6667791604995728,
+ "eval_mean_token_accuracy": 0.8082476602281843,
+ "eval_num_tokens": 822969.0,
+ "eval_runtime": 57.3807,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 320
+ },
+ {
+ "entropy": 0.7213595665991306,
+ "epoch": 0.8348680171884592,
+ "grad_norm": 0.48658841848373413,
+ "learning_rate": 0.00015627562425793356,
+ "loss": 0.6494232654571533,
+ "mean_token_accuracy": 0.811632139235735,
+ "num_tokens": 873459.0,
+ "step": 340
+ },
+ {
+ "epoch": 0.8348680171884592,
+ "eval_entropy": 0.7198746153286525,
+ "eval_loss": 0.6634471416473389,
+ "eval_mean_token_accuracy": 0.8094694914136614,
+ "eval_num_tokens": 873459.0,
+ "eval_runtime": 57.6988,
+ "eval_samples_per_second": 24.195,
+ "eval_steps_per_second": 3.033,
+ "step": 340
+ },
+ {
+ "entropy": 0.7133754774928093,
+ "epoch": 0.8839779005524862,
+ "grad_norm": 0.685793399810791,
+ "learning_rate": 0.00014790538887019238,
+ "loss": 0.642530870437622,
+ "mean_token_accuracy": 0.8122274152934551,
+ "num_tokens": 925669.0,
+ "step": 360
+ },
+ {
+ "epoch": 0.8839779005524862,
+ "eval_entropy": 0.6683637884684971,
+ "eval_loss": 0.6579923033714294,
+ "eval_mean_token_accuracy": 0.811241352558136,
+ "eval_num_tokens": 925669.0,
+ "eval_runtime": 57.8503,
+ "eval_samples_per_second": 24.131,
+ "eval_steps_per_second": 3.025,
+ "step": 360
+ },
+ {
+ "entropy": 0.7262612670660019,
+ "epoch": 0.9330877839165131,
+ "grad_norm": 0.6348966956138611,
+ "learning_rate": 0.00013923931942255727,
+ "loss": 0.656022596359253,
+ "mean_token_accuracy": 0.8072879828512669,
+ "num_tokens": 973436.0,
+ "step": 380
+ },
+ {
+ "epoch": 0.9330877839165131,
+ "eval_entropy": 0.7012172797748021,
+ "eval_loss": 0.651187002658844,
+ "eval_mean_token_accuracy": 0.8117410482679095,
+ "eval_num_tokens": 973436.0,
+ "eval_runtime": 57.1997,
+ "eval_samples_per_second": 24.406,
+ "eval_steps_per_second": 3.059,
+ "step": 380
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.635372315427584e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..8802e06314794d00a606546a836e779851ddd6bd
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/trainer_state.json
@@ -0,0 +1,76 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.09821976672805402,
+ "eval_steps": 20,
+ "global_step": 40,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1795892498050560.0,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..4c36b7d64033266b75e70a3c631d490dc5fd3d6b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/trainer_state.json
@@ -0,0 +1,454 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.9821976672805403,
+ "eval_steps": 20,
+ "global_step": 400,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ },
+ {
+ "entropy": 0.879152987897396,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 1.067221760749817,
+ "learning_rate": 0.00021473293557345806,
+ "loss": 0.807522201538086,
+ "mean_token_accuracy": 0.772921597212553,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.8607830974033901,
+ "eval_loss": 0.8036007881164551,
+ "eval_mean_token_accuracy": 0.7755460752759661,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 57.7576,
+ "eval_samples_per_second": 24.17,
+ "eval_steps_per_second": 3.03,
+ "step": 100
+ },
+ {
+ "entropy": 0.8589029781520366,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 0.7390042543411255,
+ "learning_rate": 0.0002136721067543722,
+ "loss": 0.7829637050628662,
+ "mean_token_accuracy": 0.779672809690237,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8330987681661334,
+ "eval_loss": 0.7742031216621399,
+ "eval_mean_token_accuracy": 0.785119104044778,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 57.3803,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 120
+ },
+ {
+ "entropy": 0.8094905942678452,
+ "epoch": 0.3437691835481891,
+ "grad_norm": 0.6055881381034851,
+ "learning_rate": 0.00021183381937537642,
+ "loss": 0.7396022796630859,
+ "mean_token_accuracy": 0.7867072127759457,
+ "num_tokens": 361815.0,
+ "step": 140
+ },
+ {
+ "epoch": 0.3437691835481891,
+ "eval_entropy": 0.7953300523757935,
+ "eval_loss": 0.7521983981132507,
+ "eval_mean_token_accuracy": 0.7905156462533134,
+ "eval_num_tokens": 361815.0,
+ "eval_runtime": 57.6345,
+ "eval_samples_per_second": 24.222,
+ "eval_steps_per_second": 3.036,
+ "step": 140
+ },
+ {
+ "entropy": 0.7955982603132725,
+ "epoch": 0.3928790669122161,
+ "grad_norm": 0.5252285599708557,
+ "learning_rate": 0.00020923153562932872,
+ "loss": 0.7219765186309814,
+ "mean_token_accuracy": 0.792796092480421,
+ "num_tokens": 417429.0,
+ "step": 160
+ },
+ {
+ "epoch": 0.3928790669122161,
+ "eval_entropy": 0.7623447510174343,
+ "eval_loss": 0.7298487424850464,
+ "eval_mean_token_accuracy": 0.7945844326700483,
+ "eval_num_tokens": 417429.0,
+ "eval_runtime": 57.9123,
+ "eval_samples_per_second": 24.105,
+ "eval_steps_per_second": 3.022,
+ "step": 160
+ },
+ {
+ "entropy": 0.7903043553233147,
+ "epoch": 0.4419889502762431,
+ "grad_norm": 0.6948235034942627,
+ "learning_rate": 0.00020588431262670093,
+ "loss": 0.7225021839141845,
+ "mean_token_accuracy": 0.791443546116352,
+ "num_tokens": 467817.0,
+ "step": 180
+ },
+ {
+ "epoch": 0.4419889502762431,
+ "eval_entropy": 0.790012172971453,
+ "eval_loss": 0.720366895198822,
+ "eval_mean_token_accuracy": 0.7943330574035644,
+ "eval_num_tokens": 467817.0,
+ "eval_runtime": 57.4578,
+ "eval_samples_per_second": 24.296,
+ "eval_steps_per_second": 3.046,
+ "step": 180
+ },
+ {
+ "entropy": 0.7773297265172004,
+ "epoch": 0.4910988336402701,
+ "grad_norm": 0.6160358786582947,
+ "learning_rate": 0.00020181666283606295,
+ "loss": 0.7100065231323243,
+ "mean_token_accuracy": 0.7956355385482311,
+ "num_tokens": 513773.0,
+ "step": 200
+ },
+ {
+ "epoch": 0.4910988336402701,
+ "eval_entropy": 0.783927013874054,
+ "eval_loss": 0.7202737927436829,
+ "eval_mean_token_accuracy": 0.7964657868657793,
+ "eval_num_tokens": 513773.0,
+ "eval_runtime": 57.6663,
+ "eval_samples_per_second": 24.208,
+ "eval_steps_per_second": 3.035,
+ "step": 200
+ },
+ {
+ "entropy": 0.7645334511995315,
+ "epoch": 0.5402087170042971,
+ "grad_norm": 0.6620849967002869,
+ "learning_rate": 0.00019705837457374996,
+ "loss": 0.7037524223327637,
+ "mean_token_accuracy": 0.7996554173529148,
+ "num_tokens": 565606.0,
+ "step": 220
+ },
+ {
+ "epoch": 0.5402087170042971,
+ "eval_entropy": 0.7768624458994184,
+ "eval_loss": 0.7051481604576111,
+ "eval_mean_token_accuracy": 0.7991346962111336,
+ "eval_num_tokens": 565606.0,
+ "eval_runtime": 57.7442,
+ "eval_samples_per_second": 24.176,
+ "eval_steps_per_second": 3.031,
+ "step": 220
+ },
+ {
+ "entropy": 0.7731568738818169,
+ "epoch": 0.5893186003683242,
+ "grad_norm": 0.553688645362854,
+ "learning_rate": 0.00019164429385730688,
+ "loss": 0.7003726959228516,
+ "mean_token_accuracy": 0.7977605648338795,
+ "num_tokens": 616528.0,
+ "step": 240
+ },
+ {
+ "epoch": 0.5893186003683242,
+ "eval_entropy": 0.7570131056649344,
+ "eval_loss": 0.7042225003242493,
+ "eval_mean_token_accuracy": 0.8000479991095407,
+ "eval_num_tokens": 616528.0,
+ "eval_runtime": 57.8713,
+ "eval_samples_per_second": 24.122,
+ "eval_steps_per_second": 3.024,
+ "step": 240
+ },
+ {
+ "entropy": 0.7343776792287826,
+ "epoch": 0.6384284837323512,
+ "grad_norm": 0.5576229095458984,
+ "learning_rate": 0.00018561406922024723,
+ "loss": 0.6735261440277099,
+ "mean_token_accuracy": 0.8059290319681167,
+ "num_tokens": 669440.0,
+ "step": 260
+ },
+ {
+ "epoch": 0.6384284837323512,
+ "eval_entropy": 0.7231997234480722,
+ "eval_loss": 0.6926313638687134,
+ "eval_mean_token_accuracy": 0.8031776939119611,
+ "eval_num_tokens": 669440.0,
+ "eval_runtime": 57.1692,
+ "eval_samples_per_second": 24.419,
+ "eval_steps_per_second": 3.061,
+ "step": 260
+ },
+ {
+ "entropy": 0.7326226674020291,
+ "epoch": 0.6875383670963782,
+ "grad_norm": 0.5962154865264893,
+ "learning_rate": 0.00017901186135690458,
+ "loss": 0.6677989482879638,
+ "mean_token_accuracy": 0.8022022873163224,
+ "num_tokens": 720146.0,
+ "step": 280
+ },
+ {
+ "epoch": 0.6875383670963782,
+ "eval_entropy": 0.7265675282478332,
+ "eval_loss": 0.6867054104804993,
+ "eval_mean_token_accuracy": 0.8021906747136798,
+ "eval_num_tokens": 720146.0,
+ "eval_runtime": 57.8067,
+ "eval_samples_per_second": 24.149,
+ "eval_steps_per_second": 3.027,
+ "step": 280
+ },
+ {
+ "entropy": 0.7409031428396702,
+ "epoch": 0.7366482504604052,
+ "grad_norm": 0.5716915726661682,
+ "learning_rate": 0.0001718860197237135,
+ "loss": 0.6692957878112793,
+ "mean_token_accuracy": 0.8036864519119262,
+ "num_tokens": 772624.0,
+ "step": 300
+ },
+ {
+ "epoch": 0.7366482504604052,
+ "eval_entropy": 0.7272419779641288,
+ "eval_loss": 0.6771560907363892,
+ "eval_mean_token_accuracy": 0.8062130059514727,
+ "eval_num_tokens": 772624.0,
+ "eval_runtime": 57.5942,
+ "eval_samples_per_second": 24.239,
+ "eval_steps_per_second": 3.039,
+ "step": 300
+ },
+ {
+ "entropy": 0.7243741095066071,
+ "epoch": 0.7857581338244322,
+ "grad_norm": 0.5360459089279175,
+ "learning_rate": 0.00016428872846524118,
+ "loss": 0.6628021240234375,
+ "mean_token_accuracy": 0.8050650551915168,
+ "num_tokens": 822969.0,
+ "step": 320
+ },
+ {
+ "epoch": 0.7857581338244322,
+ "eval_entropy": 0.7325100105149406,
+ "eval_loss": 0.6667791604995728,
+ "eval_mean_token_accuracy": 0.8082476602281843,
+ "eval_num_tokens": 822969.0,
+ "eval_runtime": 57.3807,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 320
+ },
+ {
+ "entropy": 0.7213595665991306,
+ "epoch": 0.8348680171884592,
+ "grad_norm": 0.48658841848373413,
+ "learning_rate": 0.00015627562425793356,
+ "loss": 0.6494232654571533,
+ "mean_token_accuracy": 0.811632139235735,
+ "num_tokens": 873459.0,
+ "step": 340
+ },
+ {
+ "epoch": 0.8348680171884592,
+ "eval_entropy": 0.7198746153286525,
+ "eval_loss": 0.6634471416473389,
+ "eval_mean_token_accuracy": 0.8094694914136614,
+ "eval_num_tokens": 873459.0,
+ "eval_runtime": 57.6988,
+ "eval_samples_per_second": 24.195,
+ "eval_steps_per_second": 3.033,
+ "step": 340
+ },
+ {
+ "entropy": 0.7133754774928093,
+ "epoch": 0.8839779005524862,
+ "grad_norm": 0.685793399810791,
+ "learning_rate": 0.00014790538887019238,
+ "loss": 0.642530870437622,
+ "mean_token_accuracy": 0.8122274152934551,
+ "num_tokens": 925669.0,
+ "step": 360
+ },
+ {
+ "epoch": 0.8839779005524862,
+ "eval_entropy": 0.6683637884684971,
+ "eval_loss": 0.6579923033714294,
+ "eval_mean_token_accuracy": 0.811241352558136,
+ "eval_num_tokens": 925669.0,
+ "eval_runtime": 57.8503,
+ "eval_samples_per_second": 24.131,
+ "eval_steps_per_second": 3.025,
+ "step": 360
+ },
+ {
+ "entropy": 0.7262612670660019,
+ "epoch": 0.9330877839165131,
+ "grad_norm": 0.6348966956138611,
+ "learning_rate": 0.00013923931942255727,
+ "loss": 0.656022596359253,
+ "mean_token_accuracy": 0.8072879828512669,
+ "num_tokens": 973436.0,
+ "step": 380
+ },
+ {
+ "epoch": 0.9330877839165131,
+ "eval_entropy": 0.7012172797748021,
+ "eval_loss": 0.651187002658844,
+ "eval_mean_token_accuracy": 0.8117410482679095,
+ "eval_num_tokens": 973436.0,
+ "eval_runtime": 57.1997,
+ "eval_samples_per_second": 24.406,
+ "eval_steps_per_second": 3.059,
+ "step": 380
+ },
+ {
+ "entropy": 0.7007582865655422,
+ "epoch": 0.9821976672805403,
+ "grad_norm": 0.4913424849510193,
+ "learning_rate": 0.0001303408794950743,
+ "loss": 0.6367915153503418,
+ "mean_token_accuracy": 0.8122764162719249,
+ "num_tokens": 1026370.0,
+ "step": 400
+ },
+ {
+ "epoch": 0.9821976672805403,
+ "eval_entropy": 0.6719393570082528,
+ "eval_loss": 0.6448967456817627,
+ "eval_mean_token_accuracy": 0.8140764743941171,
+ "eval_num_tokens": 1026370.0,
+ "eval_runtime": 57.9405,
+ "eval_samples_per_second": 24.094,
+ "eval_steps_per_second": 3.02,
+ "step": 400
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.724008424391936e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..54e42af62d5fe96454bcd84220b3cdb3f4c64013
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/trainer_state.json
@@ -0,0 +1,475 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 1.0294659300184161,
+ "eval_steps": 20,
+ "global_step": 420,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ },
+ {
+ "entropy": 0.879152987897396,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 1.067221760749817,
+ "learning_rate": 0.00021473293557345806,
+ "loss": 0.807522201538086,
+ "mean_token_accuracy": 0.772921597212553,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.8607830974033901,
+ "eval_loss": 0.8036007881164551,
+ "eval_mean_token_accuracy": 0.7755460752759661,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 57.7576,
+ "eval_samples_per_second": 24.17,
+ "eval_steps_per_second": 3.03,
+ "step": 100
+ },
+ {
+ "entropy": 0.8589029781520366,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 0.7390042543411255,
+ "learning_rate": 0.0002136721067543722,
+ "loss": 0.7829637050628662,
+ "mean_token_accuracy": 0.779672809690237,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8330987681661334,
+ "eval_loss": 0.7742031216621399,
+ "eval_mean_token_accuracy": 0.785119104044778,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 57.3803,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 120
+ },
+ {
+ "entropy": 0.8094905942678452,
+ "epoch": 0.3437691835481891,
+ "grad_norm": 0.6055881381034851,
+ "learning_rate": 0.00021183381937537642,
+ "loss": 0.7396022796630859,
+ "mean_token_accuracy": 0.7867072127759457,
+ "num_tokens": 361815.0,
+ "step": 140
+ },
+ {
+ "epoch": 0.3437691835481891,
+ "eval_entropy": 0.7953300523757935,
+ "eval_loss": 0.7521983981132507,
+ "eval_mean_token_accuracy": 0.7905156462533134,
+ "eval_num_tokens": 361815.0,
+ "eval_runtime": 57.6345,
+ "eval_samples_per_second": 24.222,
+ "eval_steps_per_second": 3.036,
+ "step": 140
+ },
+ {
+ "entropy": 0.7955982603132725,
+ "epoch": 0.3928790669122161,
+ "grad_norm": 0.5252285599708557,
+ "learning_rate": 0.00020923153562932872,
+ "loss": 0.7219765186309814,
+ "mean_token_accuracy": 0.792796092480421,
+ "num_tokens": 417429.0,
+ "step": 160
+ },
+ {
+ "epoch": 0.3928790669122161,
+ "eval_entropy": 0.7623447510174343,
+ "eval_loss": 0.7298487424850464,
+ "eval_mean_token_accuracy": 0.7945844326700483,
+ "eval_num_tokens": 417429.0,
+ "eval_runtime": 57.9123,
+ "eval_samples_per_second": 24.105,
+ "eval_steps_per_second": 3.022,
+ "step": 160
+ },
+ {
+ "entropy": 0.7903043553233147,
+ "epoch": 0.4419889502762431,
+ "grad_norm": 0.6948235034942627,
+ "learning_rate": 0.00020588431262670093,
+ "loss": 0.7225021839141845,
+ "mean_token_accuracy": 0.791443546116352,
+ "num_tokens": 467817.0,
+ "step": 180
+ },
+ {
+ "epoch": 0.4419889502762431,
+ "eval_entropy": 0.790012172971453,
+ "eval_loss": 0.720366895198822,
+ "eval_mean_token_accuracy": 0.7943330574035644,
+ "eval_num_tokens": 467817.0,
+ "eval_runtime": 57.4578,
+ "eval_samples_per_second": 24.296,
+ "eval_steps_per_second": 3.046,
+ "step": 180
+ },
+ {
+ "entropy": 0.7773297265172004,
+ "epoch": 0.4910988336402701,
+ "grad_norm": 0.6160358786582947,
+ "learning_rate": 0.00020181666283606295,
+ "loss": 0.7100065231323243,
+ "mean_token_accuracy": 0.7956355385482311,
+ "num_tokens": 513773.0,
+ "step": 200
+ },
+ {
+ "epoch": 0.4910988336402701,
+ "eval_entropy": 0.783927013874054,
+ "eval_loss": 0.7202737927436829,
+ "eval_mean_token_accuracy": 0.7964657868657793,
+ "eval_num_tokens": 513773.0,
+ "eval_runtime": 57.6663,
+ "eval_samples_per_second": 24.208,
+ "eval_steps_per_second": 3.035,
+ "step": 200
+ },
+ {
+ "entropy": 0.7645334511995315,
+ "epoch": 0.5402087170042971,
+ "grad_norm": 0.6620849967002869,
+ "learning_rate": 0.00019705837457374996,
+ "loss": 0.7037524223327637,
+ "mean_token_accuracy": 0.7996554173529148,
+ "num_tokens": 565606.0,
+ "step": 220
+ },
+ {
+ "epoch": 0.5402087170042971,
+ "eval_entropy": 0.7768624458994184,
+ "eval_loss": 0.7051481604576111,
+ "eval_mean_token_accuracy": 0.7991346962111336,
+ "eval_num_tokens": 565606.0,
+ "eval_runtime": 57.7442,
+ "eval_samples_per_second": 24.176,
+ "eval_steps_per_second": 3.031,
+ "step": 220
+ },
+ {
+ "entropy": 0.7731568738818169,
+ "epoch": 0.5893186003683242,
+ "grad_norm": 0.553688645362854,
+ "learning_rate": 0.00019164429385730688,
+ "loss": 0.7003726959228516,
+ "mean_token_accuracy": 0.7977605648338795,
+ "num_tokens": 616528.0,
+ "step": 240
+ },
+ {
+ "epoch": 0.5893186003683242,
+ "eval_entropy": 0.7570131056649344,
+ "eval_loss": 0.7042225003242493,
+ "eval_mean_token_accuracy": 0.8000479991095407,
+ "eval_num_tokens": 616528.0,
+ "eval_runtime": 57.8713,
+ "eval_samples_per_second": 24.122,
+ "eval_steps_per_second": 3.024,
+ "step": 240
+ },
+ {
+ "entropy": 0.7343776792287826,
+ "epoch": 0.6384284837323512,
+ "grad_norm": 0.5576229095458984,
+ "learning_rate": 0.00018561406922024723,
+ "loss": 0.6735261440277099,
+ "mean_token_accuracy": 0.8059290319681167,
+ "num_tokens": 669440.0,
+ "step": 260
+ },
+ {
+ "epoch": 0.6384284837323512,
+ "eval_entropy": 0.7231997234480722,
+ "eval_loss": 0.6926313638687134,
+ "eval_mean_token_accuracy": 0.8031776939119611,
+ "eval_num_tokens": 669440.0,
+ "eval_runtime": 57.1692,
+ "eval_samples_per_second": 24.419,
+ "eval_steps_per_second": 3.061,
+ "step": 260
+ },
+ {
+ "entropy": 0.7326226674020291,
+ "epoch": 0.6875383670963782,
+ "grad_norm": 0.5962154865264893,
+ "learning_rate": 0.00017901186135690458,
+ "loss": 0.6677989482879638,
+ "mean_token_accuracy": 0.8022022873163224,
+ "num_tokens": 720146.0,
+ "step": 280
+ },
+ {
+ "epoch": 0.6875383670963782,
+ "eval_entropy": 0.7265675282478332,
+ "eval_loss": 0.6867054104804993,
+ "eval_mean_token_accuracy": 0.8021906747136798,
+ "eval_num_tokens": 720146.0,
+ "eval_runtime": 57.8067,
+ "eval_samples_per_second": 24.149,
+ "eval_steps_per_second": 3.027,
+ "step": 280
+ },
+ {
+ "entropy": 0.7409031428396702,
+ "epoch": 0.7366482504604052,
+ "grad_norm": 0.5716915726661682,
+ "learning_rate": 0.0001718860197237135,
+ "loss": 0.6692957878112793,
+ "mean_token_accuracy": 0.8036864519119262,
+ "num_tokens": 772624.0,
+ "step": 300
+ },
+ {
+ "epoch": 0.7366482504604052,
+ "eval_entropy": 0.7272419779641288,
+ "eval_loss": 0.6771560907363892,
+ "eval_mean_token_accuracy": 0.8062130059514727,
+ "eval_num_tokens": 772624.0,
+ "eval_runtime": 57.5942,
+ "eval_samples_per_second": 24.239,
+ "eval_steps_per_second": 3.039,
+ "step": 300
+ },
+ {
+ "entropy": 0.7243741095066071,
+ "epoch": 0.7857581338244322,
+ "grad_norm": 0.5360459089279175,
+ "learning_rate": 0.00016428872846524118,
+ "loss": 0.6628021240234375,
+ "mean_token_accuracy": 0.8050650551915168,
+ "num_tokens": 822969.0,
+ "step": 320
+ },
+ {
+ "epoch": 0.7857581338244322,
+ "eval_entropy": 0.7325100105149406,
+ "eval_loss": 0.6667791604995728,
+ "eval_mean_token_accuracy": 0.8082476602281843,
+ "eval_num_tokens": 822969.0,
+ "eval_runtime": 57.3807,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 320
+ },
+ {
+ "entropy": 0.7213595665991306,
+ "epoch": 0.8348680171884592,
+ "grad_norm": 0.48658841848373413,
+ "learning_rate": 0.00015627562425793356,
+ "loss": 0.6494232654571533,
+ "mean_token_accuracy": 0.811632139235735,
+ "num_tokens": 873459.0,
+ "step": 340
+ },
+ {
+ "epoch": 0.8348680171884592,
+ "eval_entropy": 0.7198746153286525,
+ "eval_loss": 0.6634471416473389,
+ "eval_mean_token_accuracy": 0.8094694914136614,
+ "eval_num_tokens": 873459.0,
+ "eval_runtime": 57.6988,
+ "eval_samples_per_second": 24.195,
+ "eval_steps_per_second": 3.033,
+ "step": 340
+ },
+ {
+ "entropy": 0.7133754774928093,
+ "epoch": 0.8839779005524862,
+ "grad_norm": 0.685793399810791,
+ "learning_rate": 0.00014790538887019238,
+ "loss": 0.642530870437622,
+ "mean_token_accuracy": 0.8122274152934551,
+ "num_tokens": 925669.0,
+ "step": 360
+ },
+ {
+ "epoch": 0.8839779005524862,
+ "eval_entropy": 0.6683637884684971,
+ "eval_loss": 0.6579923033714294,
+ "eval_mean_token_accuracy": 0.811241352558136,
+ "eval_num_tokens": 925669.0,
+ "eval_runtime": 57.8503,
+ "eval_samples_per_second": 24.131,
+ "eval_steps_per_second": 3.025,
+ "step": 360
+ },
+ {
+ "entropy": 0.7262612670660019,
+ "epoch": 0.9330877839165131,
+ "grad_norm": 0.6348966956138611,
+ "learning_rate": 0.00013923931942255727,
+ "loss": 0.656022596359253,
+ "mean_token_accuracy": 0.8072879828512669,
+ "num_tokens": 973436.0,
+ "step": 380
+ },
+ {
+ "epoch": 0.9330877839165131,
+ "eval_entropy": 0.7012172797748021,
+ "eval_loss": 0.651187002658844,
+ "eval_mean_token_accuracy": 0.8117410482679095,
+ "eval_num_tokens": 973436.0,
+ "eval_runtime": 57.1997,
+ "eval_samples_per_second": 24.406,
+ "eval_steps_per_second": 3.059,
+ "step": 380
+ },
+ {
+ "entropy": 0.7007582865655422,
+ "epoch": 0.9821976672805403,
+ "grad_norm": 0.4913424849510193,
+ "learning_rate": 0.0001303408794950743,
+ "loss": 0.6367915153503418,
+ "mean_token_accuracy": 0.8122764162719249,
+ "num_tokens": 1026370.0,
+ "step": 400
+ },
+ {
+ "epoch": 0.9821976672805403,
+ "eval_entropy": 0.6719393570082528,
+ "eval_loss": 0.6448967456817627,
+ "eval_mean_token_accuracy": 0.8140764743941171,
+ "eval_num_tokens": 1026370.0,
+ "eval_runtime": 57.9405,
+ "eval_samples_per_second": 24.094,
+ "eval_steps_per_second": 3.02,
+ "step": 400
+ },
+ {
+ "entropy": 0.6647140879135627,
+ "epoch": 1.0294659300184161,
+ "grad_norm": 0.4148324429988861,
+ "learning_rate": 0.00012127523436919402,
+ "loss": 0.5996760845184326,
+ "mean_token_accuracy": 0.8205991890523341,
+ "num_tokens": 1076273.0,
+ "step": 420
+ },
+ {
+ "epoch": 1.0294659300184161,
+ "eval_entropy": 0.6531744088445391,
+ "eval_loss": 0.6474375128746033,
+ "eval_mean_token_accuracy": 0.8118055742127555,
+ "eval_num_tokens": 1076273.0,
+ "eval_runtime": 57.7161,
+ "eval_samples_per_second": 24.187,
+ "eval_steps_per_second": 3.032,
+ "step": 420
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.806675810168e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..563ff6e8594a33c55faa233ae21a7f8b45c59660
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/trainer_state.json
@@ -0,0 +1,496 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 1.0785758133824432,
+ "eval_steps": 20,
+ "global_step": 440,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ },
+ {
+ "entropy": 0.879152987897396,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 1.067221760749817,
+ "learning_rate": 0.00021473293557345806,
+ "loss": 0.807522201538086,
+ "mean_token_accuracy": 0.772921597212553,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.8607830974033901,
+ "eval_loss": 0.8036007881164551,
+ "eval_mean_token_accuracy": 0.7755460752759661,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 57.7576,
+ "eval_samples_per_second": 24.17,
+ "eval_steps_per_second": 3.03,
+ "step": 100
+ },
+ {
+ "entropy": 0.8589029781520366,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 0.7390042543411255,
+ "learning_rate": 0.0002136721067543722,
+ "loss": 0.7829637050628662,
+ "mean_token_accuracy": 0.779672809690237,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8330987681661334,
+ "eval_loss": 0.7742031216621399,
+ "eval_mean_token_accuracy": 0.785119104044778,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 57.3803,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 120
+ },
+ {
+ "entropy": 0.8094905942678452,
+ "epoch": 0.3437691835481891,
+ "grad_norm": 0.6055881381034851,
+ "learning_rate": 0.00021183381937537642,
+ "loss": 0.7396022796630859,
+ "mean_token_accuracy": 0.7867072127759457,
+ "num_tokens": 361815.0,
+ "step": 140
+ },
+ {
+ "epoch": 0.3437691835481891,
+ "eval_entropy": 0.7953300523757935,
+ "eval_loss": 0.7521983981132507,
+ "eval_mean_token_accuracy": 0.7905156462533134,
+ "eval_num_tokens": 361815.0,
+ "eval_runtime": 57.6345,
+ "eval_samples_per_second": 24.222,
+ "eval_steps_per_second": 3.036,
+ "step": 140
+ },
+ {
+ "entropy": 0.7955982603132725,
+ "epoch": 0.3928790669122161,
+ "grad_norm": 0.5252285599708557,
+ "learning_rate": 0.00020923153562932872,
+ "loss": 0.7219765186309814,
+ "mean_token_accuracy": 0.792796092480421,
+ "num_tokens": 417429.0,
+ "step": 160
+ },
+ {
+ "epoch": 0.3928790669122161,
+ "eval_entropy": 0.7623447510174343,
+ "eval_loss": 0.7298487424850464,
+ "eval_mean_token_accuracy": 0.7945844326700483,
+ "eval_num_tokens": 417429.0,
+ "eval_runtime": 57.9123,
+ "eval_samples_per_second": 24.105,
+ "eval_steps_per_second": 3.022,
+ "step": 160
+ },
+ {
+ "entropy": 0.7903043553233147,
+ "epoch": 0.4419889502762431,
+ "grad_norm": 0.6948235034942627,
+ "learning_rate": 0.00020588431262670093,
+ "loss": 0.7225021839141845,
+ "mean_token_accuracy": 0.791443546116352,
+ "num_tokens": 467817.0,
+ "step": 180
+ },
+ {
+ "epoch": 0.4419889502762431,
+ "eval_entropy": 0.790012172971453,
+ "eval_loss": 0.720366895198822,
+ "eval_mean_token_accuracy": 0.7943330574035644,
+ "eval_num_tokens": 467817.0,
+ "eval_runtime": 57.4578,
+ "eval_samples_per_second": 24.296,
+ "eval_steps_per_second": 3.046,
+ "step": 180
+ },
+ {
+ "entropy": 0.7773297265172004,
+ "epoch": 0.4910988336402701,
+ "grad_norm": 0.6160358786582947,
+ "learning_rate": 0.00020181666283606295,
+ "loss": 0.7100065231323243,
+ "mean_token_accuracy": 0.7956355385482311,
+ "num_tokens": 513773.0,
+ "step": 200
+ },
+ {
+ "epoch": 0.4910988336402701,
+ "eval_entropy": 0.783927013874054,
+ "eval_loss": 0.7202737927436829,
+ "eval_mean_token_accuracy": 0.7964657868657793,
+ "eval_num_tokens": 513773.0,
+ "eval_runtime": 57.6663,
+ "eval_samples_per_second": 24.208,
+ "eval_steps_per_second": 3.035,
+ "step": 200
+ },
+ {
+ "entropy": 0.7645334511995315,
+ "epoch": 0.5402087170042971,
+ "grad_norm": 0.6620849967002869,
+ "learning_rate": 0.00019705837457374996,
+ "loss": 0.7037524223327637,
+ "mean_token_accuracy": 0.7996554173529148,
+ "num_tokens": 565606.0,
+ "step": 220
+ },
+ {
+ "epoch": 0.5402087170042971,
+ "eval_entropy": 0.7768624458994184,
+ "eval_loss": 0.7051481604576111,
+ "eval_mean_token_accuracy": 0.7991346962111336,
+ "eval_num_tokens": 565606.0,
+ "eval_runtime": 57.7442,
+ "eval_samples_per_second": 24.176,
+ "eval_steps_per_second": 3.031,
+ "step": 220
+ },
+ {
+ "entropy": 0.7731568738818169,
+ "epoch": 0.5893186003683242,
+ "grad_norm": 0.553688645362854,
+ "learning_rate": 0.00019164429385730688,
+ "loss": 0.7003726959228516,
+ "mean_token_accuracy": 0.7977605648338795,
+ "num_tokens": 616528.0,
+ "step": 240
+ },
+ {
+ "epoch": 0.5893186003683242,
+ "eval_entropy": 0.7570131056649344,
+ "eval_loss": 0.7042225003242493,
+ "eval_mean_token_accuracy": 0.8000479991095407,
+ "eval_num_tokens": 616528.0,
+ "eval_runtime": 57.8713,
+ "eval_samples_per_second": 24.122,
+ "eval_steps_per_second": 3.024,
+ "step": 240
+ },
+ {
+ "entropy": 0.7343776792287826,
+ "epoch": 0.6384284837323512,
+ "grad_norm": 0.5576229095458984,
+ "learning_rate": 0.00018561406922024723,
+ "loss": 0.6735261440277099,
+ "mean_token_accuracy": 0.8059290319681167,
+ "num_tokens": 669440.0,
+ "step": 260
+ },
+ {
+ "epoch": 0.6384284837323512,
+ "eval_entropy": 0.7231997234480722,
+ "eval_loss": 0.6926313638687134,
+ "eval_mean_token_accuracy": 0.8031776939119611,
+ "eval_num_tokens": 669440.0,
+ "eval_runtime": 57.1692,
+ "eval_samples_per_second": 24.419,
+ "eval_steps_per_second": 3.061,
+ "step": 260
+ },
+ {
+ "entropy": 0.7326226674020291,
+ "epoch": 0.6875383670963782,
+ "grad_norm": 0.5962154865264893,
+ "learning_rate": 0.00017901186135690458,
+ "loss": 0.6677989482879638,
+ "mean_token_accuracy": 0.8022022873163224,
+ "num_tokens": 720146.0,
+ "step": 280
+ },
+ {
+ "epoch": 0.6875383670963782,
+ "eval_entropy": 0.7265675282478332,
+ "eval_loss": 0.6867054104804993,
+ "eval_mean_token_accuracy": 0.8021906747136798,
+ "eval_num_tokens": 720146.0,
+ "eval_runtime": 57.8067,
+ "eval_samples_per_second": 24.149,
+ "eval_steps_per_second": 3.027,
+ "step": 280
+ },
+ {
+ "entropy": 0.7409031428396702,
+ "epoch": 0.7366482504604052,
+ "grad_norm": 0.5716915726661682,
+ "learning_rate": 0.0001718860197237135,
+ "loss": 0.6692957878112793,
+ "mean_token_accuracy": 0.8036864519119262,
+ "num_tokens": 772624.0,
+ "step": 300
+ },
+ {
+ "epoch": 0.7366482504604052,
+ "eval_entropy": 0.7272419779641288,
+ "eval_loss": 0.6771560907363892,
+ "eval_mean_token_accuracy": 0.8062130059514727,
+ "eval_num_tokens": 772624.0,
+ "eval_runtime": 57.5942,
+ "eval_samples_per_second": 24.239,
+ "eval_steps_per_second": 3.039,
+ "step": 300
+ },
+ {
+ "entropy": 0.7243741095066071,
+ "epoch": 0.7857581338244322,
+ "grad_norm": 0.5360459089279175,
+ "learning_rate": 0.00016428872846524118,
+ "loss": 0.6628021240234375,
+ "mean_token_accuracy": 0.8050650551915168,
+ "num_tokens": 822969.0,
+ "step": 320
+ },
+ {
+ "epoch": 0.7857581338244322,
+ "eval_entropy": 0.7325100105149406,
+ "eval_loss": 0.6667791604995728,
+ "eval_mean_token_accuracy": 0.8082476602281843,
+ "eval_num_tokens": 822969.0,
+ "eval_runtime": 57.3807,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 320
+ },
+ {
+ "entropy": 0.7213595665991306,
+ "epoch": 0.8348680171884592,
+ "grad_norm": 0.48658841848373413,
+ "learning_rate": 0.00015627562425793356,
+ "loss": 0.6494232654571533,
+ "mean_token_accuracy": 0.811632139235735,
+ "num_tokens": 873459.0,
+ "step": 340
+ },
+ {
+ "epoch": 0.8348680171884592,
+ "eval_entropy": 0.7198746153286525,
+ "eval_loss": 0.6634471416473389,
+ "eval_mean_token_accuracy": 0.8094694914136614,
+ "eval_num_tokens": 873459.0,
+ "eval_runtime": 57.6988,
+ "eval_samples_per_second": 24.195,
+ "eval_steps_per_second": 3.033,
+ "step": 340
+ },
+ {
+ "entropy": 0.7133754774928093,
+ "epoch": 0.8839779005524862,
+ "grad_norm": 0.685793399810791,
+ "learning_rate": 0.00014790538887019238,
+ "loss": 0.642530870437622,
+ "mean_token_accuracy": 0.8122274152934551,
+ "num_tokens": 925669.0,
+ "step": 360
+ },
+ {
+ "epoch": 0.8839779005524862,
+ "eval_entropy": 0.6683637884684971,
+ "eval_loss": 0.6579923033714294,
+ "eval_mean_token_accuracy": 0.811241352558136,
+ "eval_num_tokens": 925669.0,
+ "eval_runtime": 57.8503,
+ "eval_samples_per_second": 24.131,
+ "eval_steps_per_second": 3.025,
+ "step": 360
+ },
+ {
+ "entropy": 0.7262612670660019,
+ "epoch": 0.9330877839165131,
+ "grad_norm": 0.6348966956138611,
+ "learning_rate": 0.00013923931942255727,
+ "loss": 0.656022596359253,
+ "mean_token_accuracy": 0.8072879828512669,
+ "num_tokens": 973436.0,
+ "step": 380
+ },
+ {
+ "epoch": 0.9330877839165131,
+ "eval_entropy": 0.7012172797748021,
+ "eval_loss": 0.651187002658844,
+ "eval_mean_token_accuracy": 0.8117410482679095,
+ "eval_num_tokens": 973436.0,
+ "eval_runtime": 57.1997,
+ "eval_samples_per_second": 24.406,
+ "eval_steps_per_second": 3.059,
+ "step": 380
+ },
+ {
+ "entropy": 0.7007582865655422,
+ "epoch": 0.9821976672805403,
+ "grad_norm": 0.4913424849510193,
+ "learning_rate": 0.0001303408794950743,
+ "loss": 0.6367915153503418,
+ "mean_token_accuracy": 0.8122764162719249,
+ "num_tokens": 1026370.0,
+ "step": 400
+ },
+ {
+ "epoch": 0.9821976672805403,
+ "eval_entropy": 0.6719393570082528,
+ "eval_loss": 0.6448967456817627,
+ "eval_mean_token_accuracy": 0.8140764743941171,
+ "eval_num_tokens": 1026370.0,
+ "eval_runtime": 57.9405,
+ "eval_samples_per_second": 24.094,
+ "eval_steps_per_second": 3.02,
+ "step": 400
+ },
+ {
+ "entropy": 0.6647140879135627,
+ "epoch": 1.0294659300184161,
+ "grad_norm": 0.4148324429988861,
+ "learning_rate": 0.00012127523436919402,
+ "loss": 0.5996760845184326,
+ "mean_token_accuracy": 0.8205991890523341,
+ "num_tokens": 1076273.0,
+ "step": 420
+ },
+ {
+ "epoch": 1.0294659300184161,
+ "eval_entropy": 0.6531744088445391,
+ "eval_loss": 0.6474375128746033,
+ "eval_mean_token_accuracy": 0.8118055742127555,
+ "eval_num_tokens": 1076273.0,
+ "eval_runtime": 57.7161,
+ "eval_samples_per_second": 24.187,
+ "eval_steps_per_second": 3.032,
+ "step": 420
+ },
+ {
+ "entropy": 0.64856808334589,
+ "epoch": 1.0785758133824432,
+ "grad_norm": 0.4203054904937744,
+ "learning_rate": 0.0001121087738077257,
+ "loss": 0.580596923828125,
+ "mean_token_accuracy": 0.8229757599532604,
+ "num_tokens": 1133367.0,
+ "step": 440
+ },
+ {
+ "epoch": 1.0785758133824432,
+ "eval_entropy": 0.6491447843824114,
+ "eval_loss": 0.6415509581565857,
+ "eval_mean_token_accuracy": 0.8132358625956944,
+ "eval_num_tokens": 1133367.0,
+ "eval_runtime": 57.0688,
+ "eval_samples_per_second": 24.462,
+ "eval_steps_per_second": 3.066,
+ "step": 440
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.899248419581888e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..fd518b30705bc32a44f7998f1590f93ba1593eb8
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/trainer_state.json
@@ -0,0 +1,517 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 1.1276856967464703,
+ "eval_steps": 20,
+ "global_step": 460,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ },
+ {
+ "entropy": 0.879152987897396,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 1.067221760749817,
+ "learning_rate": 0.00021473293557345806,
+ "loss": 0.807522201538086,
+ "mean_token_accuracy": 0.772921597212553,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.8607830974033901,
+ "eval_loss": 0.8036007881164551,
+ "eval_mean_token_accuracy": 0.7755460752759661,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 57.7576,
+ "eval_samples_per_second": 24.17,
+ "eval_steps_per_second": 3.03,
+ "step": 100
+ },
+ {
+ "entropy": 0.8589029781520366,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 0.7390042543411255,
+ "learning_rate": 0.0002136721067543722,
+ "loss": 0.7829637050628662,
+ "mean_token_accuracy": 0.779672809690237,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8330987681661334,
+ "eval_loss": 0.7742031216621399,
+ "eval_mean_token_accuracy": 0.785119104044778,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 57.3803,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 120
+ },
+ {
+ "entropy": 0.8094905942678452,
+ "epoch": 0.3437691835481891,
+ "grad_norm": 0.6055881381034851,
+ "learning_rate": 0.00021183381937537642,
+ "loss": 0.7396022796630859,
+ "mean_token_accuracy": 0.7867072127759457,
+ "num_tokens": 361815.0,
+ "step": 140
+ },
+ {
+ "epoch": 0.3437691835481891,
+ "eval_entropy": 0.7953300523757935,
+ "eval_loss": 0.7521983981132507,
+ "eval_mean_token_accuracy": 0.7905156462533134,
+ "eval_num_tokens": 361815.0,
+ "eval_runtime": 57.6345,
+ "eval_samples_per_second": 24.222,
+ "eval_steps_per_second": 3.036,
+ "step": 140
+ },
+ {
+ "entropy": 0.7955982603132725,
+ "epoch": 0.3928790669122161,
+ "grad_norm": 0.5252285599708557,
+ "learning_rate": 0.00020923153562932872,
+ "loss": 0.7219765186309814,
+ "mean_token_accuracy": 0.792796092480421,
+ "num_tokens": 417429.0,
+ "step": 160
+ },
+ {
+ "epoch": 0.3928790669122161,
+ "eval_entropy": 0.7623447510174343,
+ "eval_loss": 0.7298487424850464,
+ "eval_mean_token_accuracy": 0.7945844326700483,
+ "eval_num_tokens": 417429.0,
+ "eval_runtime": 57.9123,
+ "eval_samples_per_second": 24.105,
+ "eval_steps_per_second": 3.022,
+ "step": 160
+ },
+ {
+ "entropy": 0.7903043553233147,
+ "epoch": 0.4419889502762431,
+ "grad_norm": 0.6948235034942627,
+ "learning_rate": 0.00020588431262670093,
+ "loss": 0.7225021839141845,
+ "mean_token_accuracy": 0.791443546116352,
+ "num_tokens": 467817.0,
+ "step": 180
+ },
+ {
+ "epoch": 0.4419889502762431,
+ "eval_entropy": 0.790012172971453,
+ "eval_loss": 0.720366895198822,
+ "eval_mean_token_accuracy": 0.7943330574035644,
+ "eval_num_tokens": 467817.0,
+ "eval_runtime": 57.4578,
+ "eval_samples_per_second": 24.296,
+ "eval_steps_per_second": 3.046,
+ "step": 180
+ },
+ {
+ "entropy": 0.7773297265172004,
+ "epoch": 0.4910988336402701,
+ "grad_norm": 0.6160358786582947,
+ "learning_rate": 0.00020181666283606295,
+ "loss": 0.7100065231323243,
+ "mean_token_accuracy": 0.7956355385482311,
+ "num_tokens": 513773.0,
+ "step": 200
+ },
+ {
+ "epoch": 0.4910988336402701,
+ "eval_entropy": 0.783927013874054,
+ "eval_loss": 0.7202737927436829,
+ "eval_mean_token_accuracy": 0.7964657868657793,
+ "eval_num_tokens": 513773.0,
+ "eval_runtime": 57.6663,
+ "eval_samples_per_second": 24.208,
+ "eval_steps_per_second": 3.035,
+ "step": 200
+ },
+ {
+ "entropy": 0.7645334511995315,
+ "epoch": 0.5402087170042971,
+ "grad_norm": 0.6620849967002869,
+ "learning_rate": 0.00019705837457374996,
+ "loss": 0.7037524223327637,
+ "mean_token_accuracy": 0.7996554173529148,
+ "num_tokens": 565606.0,
+ "step": 220
+ },
+ {
+ "epoch": 0.5402087170042971,
+ "eval_entropy": 0.7768624458994184,
+ "eval_loss": 0.7051481604576111,
+ "eval_mean_token_accuracy": 0.7991346962111336,
+ "eval_num_tokens": 565606.0,
+ "eval_runtime": 57.7442,
+ "eval_samples_per_second": 24.176,
+ "eval_steps_per_second": 3.031,
+ "step": 220
+ },
+ {
+ "entropy": 0.7731568738818169,
+ "epoch": 0.5893186003683242,
+ "grad_norm": 0.553688645362854,
+ "learning_rate": 0.00019164429385730688,
+ "loss": 0.7003726959228516,
+ "mean_token_accuracy": 0.7977605648338795,
+ "num_tokens": 616528.0,
+ "step": 240
+ },
+ {
+ "epoch": 0.5893186003683242,
+ "eval_entropy": 0.7570131056649344,
+ "eval_loss": 0.7042225003242493,
+ "eval_mean_token_accuracy": 0.8000479991095407,
+ "eval_num_tokens": 616528.0,
+ "eval_runtime": 57.8713,
+ "eval_samples_per_second": 24.122,
+ "eval_steps_per_second": 3.024,
+ "step": 240
+ },
+ {
+ "entropy": 0.7343776792287826,
+ "epoch": 0.6384284837323512,
+ "grad_norm": 0.5576229095458984,
+ "learning_rate": 0.00018561406922024723,
+ "loss": 0.6735261440277099,
+ "mean_token_accuracy": 0.8059290319681167,
+ "num_tokens": 669440.0,
+ "step": 260
+ },
+ {
+ "epoch": 0.6384284837323512,
+ "eval_entropy": 0.7231997234480722,
+ "eval_loss": 0.6926313638687134,
+ "eval_mean_token_accuracy": 0.8031776939119611,
+ "eval_num_tokens": 669440.0,
+ "eval_runtime": 57.1692,
+ "eval_samples_per_second": 24.419,
+ "eval_steps_per_second": 3.061,
+ "step": 260
+ },
+ {
+ "entropy": 0.7326226674020291,
+ "epoch": 0.6875383670963782,
+ "grad_norm": 0.5962154865264893,
+ "learning_rate": 0.00017901186135690458,
+ "loss": 0.6677989482879638,
+ "mean_token_accuracy": 0.8022022873163224,
+ "num_tokens": 720146.0,
+ "step": 280
+ },
+ {
+ "epoch": 0.6875383670963782,
+ "eval_entropy": 0.7265675282478332,
+ "eval_loss": 0.6867054104804993,
+ "eval_mean_token_accuracy": 0.8021906747136798,
+ "eval_num_tokens": 720146.0,
+ "eval_runtime": 57.8067,
+ "eval_samples_per_second": 24.149,
+ "eval_steps_per_second": 3.027,
+ "step": 280
+ },
+ {
+ "entropy": 0.7409031428396702,
+ "epoch": 0.7366482504604052,
+ "grad_norm": 0.5716915726661682,
+ "learning_rate": 0.0001718860197237135,
+ "loss": 0.6692957878112793,
+ "mean_token_accuracy": 0.8036864519119262,
+ "num_tokens": 772624.0,
+ "step": 300
+ },
+ {
+ "epoch": 0.7366482504604052,
+ "eval_entropy": 0.7272419779641288,
+ "eval_loss": 0.6771560907363892,
+ "eval_mean_token_accuracy": 0.8062130059514727,
+ "eval_num_tokens": 772624.0,
+ "eval_runtime": 57.5942,
+ "eval_samples_per_second": 24.239,
+ "eval_steps_per_second": 3.039,
+ "step": 300
+ },
+ {
+ "entropy": 0.7243741095066071,
+ "epoch": 0.7857581338244322,
+ "grad_norm": 0.5360459089279175,
+ "learning_rate": 0.00016428872846524118,
+ "loss": 0.6628021240234375,
+ "mean_token_accuracy": 0.8050650551915168,
+ "num_tokens": 822969.0,
+ "step": 320
+ },
+ {
+ "epoch": 0.7857581338244322,
+ "eval_entropy": 0.7325100105149406,
+ "eval_loss": 0.6667791604995728,
+ "eval_mean_token_accuracy": 0.8082476602281843,
+ "eval_num_tokens": 822969.0,
+ "eval_runtime": 57.3807,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 320
+ },
+ {
+ "entropy": 0.7213595665991306,
+ "epoch": 0.8348680171884592,
+ "grad_norm": 0.48658841848373413,
+ "learning_rate": 0.00015627562425793356,
+ "loss": 0.6494232654571533,
+ "mean_token_accuracy": 0.811632139235735,
+ "num_tokens": 873459.0,
+ "step": 340
+ },
+ {
+ "epoch": 0.8348680171884592,
+ "eval_entropy": 0.7198746153286525,
+ "eval_loss": 0.6634471416473389,
+ "eval_mean_token_accuracy": 0.8094694914136614,
+ "eval_num_tokens": 873459.0,
+ "eval_runtime": 57.6988,
+ "eval_samples_per_second": 24.195,
+ "eval_steps_per_second": 3.033,
+ "step": 340
+ },
+ {
+ "entropy": 0.7133754774928093,
+ "epoch": 0.8839779005524862,
+ "grad_norm": 0.685793399810791,
+ "learning_rate": 0.00014790538887019238,
+ "loss": 0.642530870437622,
+ "mean_token_accuracy": 0.8122274152934551,
+ "num_tokens": 925669.0,
+ "step": 360
+ },
+ {
+ "epoch": 0.8839779005524862,
+ "eval_entropy": 0.6683637884684971,
+ "eval_loss": 0.6579923033714294,
+ "eval_mean_token_accuracy": 0.811241352558136,
+ "eval_num_tokens": 925669.0,
+ "eval_runtime": 57.8503,
+ "eval_samples_per_second": 24.131,
+ "eval_steps_per_second": 3.025,
+ "step": 360
+ },
+ {
+ "entropy": 0.7262612670660019,
+ "epoch": 0.9330877839165131,
+ "grad_norm": 0.6348966956138611,
+ "learning_rate": 0.00013923931942255727,
+ "loss": 0.656022596359253,
+ "mean_token_accuracy": 0.8072879828512669,
+ "num_tokens": 973436.0,
+ "step": 380
+ },
+ {
+ "epoch": 0.9330877839165131,
+ "eval_entropy": 0.7012172797748021,
+ "eval_loss": 0.651187002658844,
+ "eval_mean_token_accuracy": 0.8117410482679095,
+ "eval_num_tokens": 973436.0,
+ "eval_runtime": 57.1997,
+ "eval_samples_per_second": 24.406,
+ "eval_steps_per_second": 3.059,
+ "step": 380
+ },
+ {
+ "entropy": 0.7007582865655422,
+ "epoch": 0.9821976672805403,
+ "grad_norm": 0.4913424849510193,
+ "learning_rate": 0.0001303408794950743,
+ "loss": 0.6367915153503418,
+ "mean_token_accuracy": 0.8122764162719249,
+ "num_tokens": 1026370.0,
+ "step": 400
+ },
+ {
+ "epoch": 0.9821976672805403,
+ "eval_entropy": 0.6719393570082528,
+ "eval_loss": 0.6448967456817627,
+ "eval_mean_token_accuracy": 0.8140764743941171,
+ "eval_num_tokens": 1026370.0,
+ "eval_runtime": 57.9405,
+ "eval_samples_per_second": 24.094,
+ "eval_steps_per_second": 3.02,
+ "step": 400
+ },
+ {
+ "entropy": 0.6647140879135627,
+ "epoch": 1.0294659300184161,
+ "grad_norm": 0.4148324429988861,
+ "learning_rate": 0.00012127523436919402,
+ "loss": 0.5996760845184326,
+ "mean_token_accuracy": 0.8205991890523341,
+ "num_tokens": 1076273.0,
+ "step": 420
+ },
+ {
+ "epoch": 1.0294659300184161,
+ "eval_entropy": 0.6531744088445391,
+ "eval_loss": 0.6474375128746033,
+ "eval_mean_token_accuracy": 0.8118055742127555,
+ "eval_num_tokens": 1076273.0,
+ "eval_runtime": 57.7161,
+ "eval_samples_per_second": 24.187,
+ "eval_steps_per_second": 3.032,
+ "step": 420
+ },
+ {
+ "entropy": 0.64856808334589,
+ "epoch": 1.0785758133824432,
+ "grad_norm": 0.4203054904937744,
+ "learning_rate": 0.0001121087738077257,
+ "loss": 0.580596923828125,
+ "mean_token_accuracy": 0.8229757599532604,
+ "num_tokens": 1133367.0,
+ "step": 440
+ },
+ {
+ "epoch": 1.0785758133824432,
+ "eval_entropy": 0.6491447843824114,
+ "eval_loss": 0.6415509581565857,
+ "eval_mean_token_accuracy": 0.8132358625956944,
+ "eval_num_tokens": 1133367.0,
+ "eval_runtime": 57.0688,
+ "eval_samples_per_second": 24.462,
+ "eval_steps_per_second": 3.066,
+ "step": 440
+ },
+ {
+ "entropy": 0.6387462258338928,
+ "epoch": 1.1276856967464703,
+ "grad_norm": 0.4547317326068878,
+ "learning_rate": 0.0001029086258676389,
+ "loss": 0.5705233097076416,
+ "mean_token_accuracy": 0.8258621700108051,
+ "num_tokens": 1186585.0,
+ "step": 460
+ },
+ {
+ "epoch": 1.1276856967464703,
+ "eval_entropy": 0.653609961441585,
+ "eval_loss": 0.6332530975341797,
+ "eval_mean_token_accuracy": 0.8161680047852652,
+ "eval_num_tokens": 1186585.0,
+ "eval_runtime": 57.3472,
+ "eval_samples_per_second": 24.343,
+ "eval_steps_per_second": 3.052,
+ "step": 460
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.9901805283632e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e5fd77ce1f980ae026dca5194704c78fa7fdd8f6
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/trainer_state.json
@@ -0,0 +1,538 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 1.1767955801104972,
+ "eval_steps": 20,
+ "global_step": 480,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ },
+ {
+ "entropy": 0.879152987897396,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 1.067221760749817,
+ "learning_rate": 0.00021473293557345806,
+ "loss": 0.807522201538086,
+ "mean_token_accuracy": 0.772921597212553,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.8607830974033901,
+ "eval_loss": 0.8036007881164551,
+ "eval_mean_token_accuracy": 0.7755460752759661,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 57.7576,
+ "eval_samples_per_second": 24.17,
+ "eval_steps_per_second": 3.03,
+ "step": 100
+ },
+ {
+ "entropy": 0.8589029781520366,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 0.7390042543411255,
+ "learning_rate": 0.0002136721067543722,
+ "loss": 0.7829637050628662,
+ "mean_token_accuracy": 0.779672809690237,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8330987681661334,
+ "eval_loss": 0.7742031216621399,
+ "eval_mean_token_accuracy": 0.785119104044778,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 57.3803,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 120
+ },
+ {
+ "entropy": 0.8094905942678452,
+ "epoch": 0.3437691835481891,
+ "grad_norm": 0.6055881381034851,
+ "learning_rate": 0.00021183381937537642,
+ "loss": 0.7396022796630859,
+ "mean_token_accuracy": 0.7867072127759457,
+ "num_tokens": 361815.0,
+ "step": 140
+ },
+ {
+ "epoch": 0.3437691835481891,
+ "eval_entropy": 0.7953300523757935,
+ "eval_loss": 0.7521983981132507,
+ "eval_mean_token_accuracy": 0.7905156462533134,
+ "eval_num_tokens": 361815.0,
+ "eval_runtime": 57.6345,
+ "eval_samples_per_second": 24.222,
+ "eval_steps_per_second": 3.036,
+ "step": 140
+ },
+ {
+ "entropy": 0.7955982603132725,
+ "epoch": 0.3928790669122161,
+ "grad_norm": 0.5252285599708557,
+ "learning_rate": 0.00020923153562932872,
+ "loss": 0.7219765186309814,
+ "mean_token_accuracy": 0.792796092480421,
+ "num_tokens": 417429.0,
+ "step": 160
+ },
+ {
+ "epoch": 0.3928790669122161,
+ "eval_entropy": 0.7623447510174343,
+ "eval_loss": 0.7298487424850464,
+ "eval_mean_token_accuracy": 0.7945844326700483,
+ "eval_num_tokens": 417429.0,
+ "eval_runtime": 57.9123,
+ "eval_samples_per_second": 24.105,
+ "eval_steps_per_second": 3.022,
+ "step": 160
+ },
+ {
+ "entropy": 0.7903043553233147,
+ "epoch": 0.4419889502762431,
+ "grad_norm": 0.6948235034942627,
+ "learning_rate": 0.00020588431262670093,
+ "loss": 0.7225021839141845,
+ "mean_token_accuracy": 0.791443546116352,
+ "num_tokens": 467817.0,
+ "step": 180
+ },
+ {
+ "epoch": 0.4419889502762431,
+ "eval_entropy": 0.790012172971453,
+ "eval_loss": 0.720366895198822,
+ "eval_mean_token_accuracy": 0.7943330574035644,
+ "eval_num_tokens": 467817.0,
+ "eval_runtime": 57.4578,
+ "eval_samples_per_second": 24.296,
+ "eval_steps_per_second": 3.046,
+ "step": 180
+ },
+ {
+ "entropy": 0.7773297265172004,
+ "epoch": 0.4910988336402701,
+ "grad_norm": 0.6160358786582947,
+ "learning_rate": 0.00020181666283606295,
+ "loss": 0.7100065231323243,
+ "mean_token_accuracy": 0.7956355385482311,
+ "num_tokens": 513773.0,
+ "step": 200
+ },
+ {
+ "epoch": 0.4910988336402701,
+ "eval_entropy": 0.783927013874054,
+ "eval_loss": 0.7202737927436829,
+ "eval_mean_token_accuracy": 0.7964657868657793,
+ "eval_num_tokens": 513773.0,
+ "eval_runtime": 57.6663,
+ "eval_samples_per_second": 24.208,
+ "eval_steps_per_second": 3.035,
+ "step": 200
+ },
+ {
+ "entropy": 0.7645334511995315,
+ "epoch": 0.5402087170042971,
+ "grad_norm": 0.6620849967002869,
+ "learning_rate": 0.00019705837457374996,
+ "loss": 0.7037524223327637,
+ "mean_token_accuracy": 0.7996554173529148,
+ "num_tokens": 565606.0,
+ "step": 220
+ },
+ {
+ "epoch": 0.5402087170042971,
+ "eval_entropy": 0.7768624458994184,
+ "eval_loss": 0.7051481604576111,
+ "eval_mean_token_accuracy": 0.7991346962111336,
+ "eval_num_tokens": 565606.0,
+ "eval_runtime": 57.7442,
+ "eval_samples_per_second": 24.176,
+ "eval_steps_per_second": 3.031,
+ "step": 220
+ },
+ {
+ "entropy": 0.7731568738818169,
+ "epoch": 0.5893186003683242,
+ "grad_norm": 0.553688645362854,
+ "learning_rate": 0.00019164429385730688,
+ "loss": 0.7003726959228516,
+ "mean_token_accuracy": 0.7977605648338795,
+ "num_tokens": 616528.0,
+ "step": 240
+ },
+ {
+ "epoch": 0.5893186003683242,
+ "eval_entropy": 0.7570131056649344,
+ "eval_loss": 0.7042225003242493,
+ "eval_mean_token_accuracy": 0.8000479991095407,
+ "eval_num_tokens": 616528.0,
+ "eval_runtime": 57.8713,
+ "eval_samples_per_second": 24.122,
+ "eval_steps_per_second": 3.024,
+ "step": 240
+ },
+ {
+ "entropy": 0.7343776792287826,
+ "epoch": 0.6384284837323512,
+ "grad_norm": 0.5576229095458984,
+ "learning_rate": 0.00018561406922024723,
+ "loss": 0.6735261440277099,
+ "mean_token_accuracy": 0.8059290319681167,
+ "num_tokens": 669440.0,
+ "step": 260
+ },
+ {
+ "epoch": 0.6384284837323512,
+ "eval_entropy": 0.7231997234480722,
+ "eval_loss": 0.6926313638687134,
+ "eval_mean_token_accuracy": 0.8031776939119611,
+ "eval_num_tokens": 669440.0,
+ "eval_runtime": 57.1692,
+ "eval_samples_per_second": 24.419,
+ "eval_steps_per_second": 3.061,
+ "step": 260
+ },
+ {
+ "entropy": 0.7326226674020291,
+ "epoch": 0.6875383670963782,
+ "grad_norm": 0.5962154865264893,
+ "learning_rate": 0.00017901186135690458,
+ "loss": 0.6677989482879638,
+ "mean_token_accuracy": 0.8022022873163224,
+ "num_tokens": 720146.0,
+ "step": 280
+ },
+ {
+ "epoch": 0.6875383670963782,
+ "eval_entropy": 0.7265675282478332,
+ "eval_loss": 0.6867054104804993,
+ "eval_mean_token_accuracy": 0.8021906747136798,
+ "eval_num_tokens": 720146.0,
+ "eval_runtime": 57.8067,
+ "eval_samples_per_second": 24.149,
+ "eval_steps_per_second": 3.027,
+ "step": 280
+ },
+ {
+ "entropy": 0.7409031428396702,
+ "epoch": 0.7366482504604052,
+ "grad_norm": 0.5716915726661682,
+ "learning_rate": 0.0001718860197237135,
+ "loss": 0.6692957878112793,
+ "mean_token_accuracy": 0.8036864519119262,
+ "num_tokens": 772624.0,
+ "step": 300
+ },
+ {
+ "epoch": 0.7366482504604052,
+ "eval_entropy": 0.7272419779641288,
+ "eval_loss": 0.6771560907363892,
+ "eval_mean_token_accuracy": 0.8062130059514727,
+ "eval_num_tokens": 772624.0,
+ "eval_runtime": 57.5942,
+ "eval_samples_per_second": 24.239,
+ "eval_steps_per_second": 3.039,
+ "step": 300
+ },
+ {
+ "entropy": 0.7243741095066071,
+ "epoch": 0.7857581338244322,
+ "grad_norm": 0.5360459089279175,
+ "learning_rate": 0.00016428872846524118,
+ "loss": 0.6628021240234375,
+ "mean_token_accuracy": 0.8050650551915168,
+ "num_tokens": 822969.0,
+ "step": 320
+ },
+ {
+ "epoch": 0.7857581338244322,
+ "eval_entropy": 0.7325100105149406,
+ "eval_loss": 0.6667791604995728,
+ "eval_mean_token_accuracy": 0.8082476602281843,
+ "eval_num_tokens": 822969.0,
+ "eval_runtime": 57.3807,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 320
+ },
+ {
+ "entropy": 0.7213595665991306,
+ "epoch": 0.8348680171884592,
+ "grad_norm": 0.48658841848373413,
+ "learning_rate": 0.00015627562425793356,
+ "loss": 0.6494232654571533,
+ "mean_token_accuracy": 0.811632139235735,
+ "num_tokens": 873459.0,
+ "step": 340
+ },
+ {
+ "epoch": 0.8348680171884592,
+ "eval_entropy": 0.7198746153286525,
+ "eval_loss": 0.6634471416473389,
+ "eval_mean_token_accuracy": 0.8094694914136614,
+ "eval_num_tokens": 873459.0,
+ "eval_runtime": 57.6988,
+ "eval_samples_per_second": 24.195,
+ "eval_steps_per_second": 3.033,
+ "step": 340
+ },
+ {
+ "entropy": 0.7133754774928093,
+ "epoch": 0.8839779005524862,
+ "grad_norm": 0.685793399810791,
+ "learning_rate": 0.00014790538887019238,
+ "loss": 0.642530870437622,
+ "mean_token_accuracy": 0.8122274152934551,
+ "num_tokens": 925669.0,
+ "step": 360
+ },
+ {
+ "epoch": 0.8839779005524862,
+ "eval_entropy": 0.6683637884684971,
+ "eval_loss": 0.6579923033714294,
+ "eval_mean_token_accuracy": 0.811241352558136,
+ "eval_num_tokens": 925669.0,
+ "eval_runtime": 57.8503,
+ "eval_samples_per_second": 24.131,
+ "eval_steps_per_second": 3.025,
+ "step": 360
+ },
+ {
+ "entropy": 0.7262612670660019,
+ "epoch": 0.9330877839165131,
+ "grad_norm": 0.6348966956138611,
+ "learning_rate": 0.00013923931942255727,
+ "loss": 0.656022596359253,
+ "mean_token_accuracy": 0.8072879828512669,
+ "num_tokens": 973436.0,
+ "step": 380
+ },
+ {
+ "epoch": 0.9330877839165131,
+ "eval_entropy": 0.7012172797748021,
+ "eval_loss": 0.651187002658844,
+ "eval_mean_token_accuracy": 0.8117410482679095,
+ "eval_num_tokens": 973436.0,
+ "eval_runtime": 57.1997,
+ "eval_samples_per_second": 24.406,
+ "eval_steps_per_second": 3.059,
+ "step": 380
+ },
+ {
+ "entropy": 0.7007582865655422,
+ "epoch": 0.9821976672805403,
+ "grad_norm": 0.4913424849510193,
+ "learning_rate": 0.0001303408794950743,
+ "loss": 0.6367915153503418,
+ "mean_token_accuracy": 0.8122764162719249,
+ "num_tokens": 1026370.0,
+ "step": 400
+ },
+ {
+ "epoch": 0.9821976672805403,
+ "eval_entropy": 0.6719393570082528,
+ "eval_loss": 0.6448967456817627,
+ "eval_mean_token_accuracy": 0.8140764743941171,
+ "eval_num_tokens": 1026370.0,
+ "eval_runtime": 57.9405,
+ "eval_samples_per_second": 24.094,
+ "eval_steps_per_second": 3.02,
+ "step": 400
+ },
+ {
+ "entropy": 0.6647140879135627,
+ "epoch": 1.0294659300184161,
+ "grad_norm": 0.4148324429988861,
+ "learning_rate": 0.00012127523436919402,
+ "loss": 0.5996760845184326,
+ "mean_token_accuracy": 0.8205991890523341,
+ "num_tokens": 1076273.0,
+ "step": 420
+ },
+ {
+ "epoch": 1.0294659300184161,
+ "eval_entropy": 0.6531744088445391,
+ "eval_loss": 0.6474375128746033,
+ "eval_mean_token_accuracy": 0.8118055742127555,
+ "eval_num_tokens": 1076273.0,
+ "eval_runtime": 57.7161,
+ "eval_samples_per_second": 24.187,
+ "eval_steps_per_second": 3.032,
+ "step": 420
+ },
+ {
+ "entropy": 0.64856808334589,
+ "epoch": 1.0785758133824432,
+ "grad_norm": 0.4203054904937744,
+ "learning_rate": 0.0001121087738077257,
+ "loss": 0.580596923828125,
+ "mean_token_accuracy": 0.8229757599532604,
+ "num_tokens": 1133367.0,
+ "step": 440
+ },
+ {
+ "epoch": 1.0785758133824432,
+ "eval_entropy": 0.6491447843824114,
+ "eval_loss": 0.6415509581565857,
+ "eval_mean_token_accuracy": 0.8132358625956944,
+ "eval_num_tokens": 1133367.0,
+ "eval_runtime": 57.0688,
+ "eval_samples_per_second": 24.462,
+ "eval_steps_per_second": 3.066,
+ "step": 440
+ },
+ {
+ "entropy": 0.6387462258338928,
+ "epoch": 1.1276856967464703,
+ "grad_norm": 0.4547317326068878,
+ "learning_rate": 0.0001029086258676389,
+ "loss": 0.5705233097076416,
+ "mean_token_accuracy": 0.8258621700108051,
+ "num_tokens": 1186585.0,
+ "step": 460
+ },
+ {
+ "epoch": 1.1276856967464703,
+ "eval_entropy": 0.653609961441585,
+ "eval_loss": 0.6332530975341797,
+ "eval_mean_token_accuracy": 0.8161680047852652,
+ "eval_num_tokens": 1186585.0,
+ "eval_runtime": 57.3472,
+ "eval_samples_per_second": 24.343,
+ "eval_steps_per_second": 3.052,
+ "step": 460
+ },
+ {
+ "entropy": 0.6427005555480718,
+ "epoch": 1.1767955801104972,
+ "grad_norm": 0.4293302297592163,
+ "learning_rate": 9.374216530617052e-05,
+ "loss": 0.5639504432678223,
+ "mean_token_accuracy": 0.8271843157708645,
+ "num_tokens": 1235457.0,
+ "step": 480
+ },
+ {
+ "epoch": 1.1767955801104972,
+ "eval_entropy": 0.6493809008598328,
+ "eval_loss": 0.6308221817016602,
+ "eval_mean_token_accuracy": 0.8174078563281468,
+ "eval_num_tokens": 1235457.0,
+ "eval_runtime": 57.7158,
+ "eval_samples_per_second": 24.187,
+ "eval_steps_per_second": 3.032,
+ "step": 480
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 2.07315200266464e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..227d8ad777734a57ddcc9813830b471fe12f4610
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/trainer_state.json
@@ -0,0 +1,559 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 1.2259054634745243,
+ "eval_steps": 20,
+ "global_step": 500,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ },
+ {
+ "entropy": 0.879152987897396,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 1.067221760749817,
+ "learning_rate": 0.00021473293557345806,
+ "loss": 0.807522201538086,
+ "mean_token_accuracy": 0.772921597212553,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.8607830974033901,
+ "eval_loss": 0.8036007881164551,
+ "eval_mean_token_accuracy": 0.7755460752759661,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 57.7576,
+ "eval_samples_per_second": 24.17,
+ "eval_steps_per_second": 3.03,
+ "step": 100
+ },
+ {
+ "entropy": 0.8589029781520366,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 0.7390042543411255,
+ "learning_rate": 0.0002136721067543722,
+ "loss": 0.7829637050628662,
+ "mean_token_accuracy": 0.779672809690237,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8330987681661334,
+ "eval_loss": 0.7742031216621399,
+ "eval_mean_token_accuracy": 0.785119104044778,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 57.3803,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 120
+ },
+ {
+ "entropy": 0.8094905942678452,
+ "epoch": 0.3437691835481891,
+ "grad_norm": 0.6055881381034851,
+ "learning_rate": 0.00021183381937537642,
+ "loss": 0.7396022796630859,
+ "mean_token_accuracy": 0.7867072127759457,
+ "num_tokens": 361815.0,
+ "step": 140
+ },
+ {
+ "epoch": 0.3437691835481891,
+ "eval_entropy": 0.7953300523757935,
+ "eval_loss": 0.7521983981132507,
+ "eval_mean_token_accuracy": 0.7905156462533134,
+ "eval_num_tokens": 361815.0,
+ "eval_runtime": 57.6345,
+ "eval_samples_per_second": 24.222,
+ "eval_steps_per_second": 3.036,
+ "step": 140
+ },
+ {
+ "entropy": 0.7955982603132725,
+ "epoch": 0.3928790669122161,
+ "grad_norm": 0.5252285599708557,
+ "learning_rate": 0.00020923153562932872,
+ "loss": 0.7219765186309814,
+ "mean_token_accuracy": 0.792796092480421,
+ "num_tokens": 417429.0,
+ "step": 160
+ },
+ {
+ "epoch": 0.3928790669122161,
+ "eval_entropy": 0.7623447510174343,
+ "eval_loss": 0.7298487424850464,
+ "eval_mean_token_accuracy": 0.7945844326700483,
+ "eval_num_tokens": 417429.0,
+ "eval_runtime": 57.9123,
+ "eval_samples_per_second": 24.105,
+ "eval_steps_per_second": 3.022,
+ "step": 160
+ },
+ {
+ "entropy": 0.7903043553233147,
+ "epoch": 0.4419889502762431,
+ "grad_norm": 0.6948235034942627,
+ "learning_rate": 0.00020588431262670093,
+ "loss": 0.7225021839141845,
+ "mean_token_accuracy": 0.791443546116352,
+ "num_tokens": 467817.0,
+ "step": 180
+ },
+ {
+ "epoch": 0.4419889502762431,
+ "eval_entropy": 0.790012172971453,
+ "eval_loss": 0.720366895198822,
+ "eval_mean_token_accuracy": 0.7943330574035644,
+ "eval_num_tokens": 467817.0,
+ "eval_runtime": 57.4578,
+ "eval_samples_per_second": 24.296,
+ "eval_steps_per_second": 3.046,
+ "step": 180
+ },
+ {
+ "entropy": 0.7773297265172004,
+ "epoch": 0.4910988336402701,
+ "grad_norm": 0.6160358786582947,
+ "learning_rate": 0.00020181666283606295,
+ "loss": 0.7100065231323243,
+ "mean_token_accuracy": 0.7956355385482311,
+ "num_tokens": 513773.0,
+ "step": 200
+ },
+ {
+ "epoch": 0.4910988336402701,
+ "eval_entropy": 0.783927013874054,
+ "eval_loss": 0.7202737927436829,
+ "eval_mean_token_accuracy": 0.7964657868657793,
+ "eval_num_tokens": 513773.0,
+ "eval_runtime": 57.6663,
+ "eval_samples_per_second": 24.208,
+ "eval_steps_per_second": 3.035,
+ "step": 200
+ },
+ {
+ "entropy": 0.7645334511995315,
+ "epoch": 0.5402087170042971,
+ "grad_norm": 0.6620849967002869,
+ "learning_rate": 0.00019705837457374996,
+ "loss": 0.7037524223327637,
+ "mean_token_accuracy": 0.7996554173529148,
+ "num_tokens": 565606.0,
+ "step": 220
+ },
+ {
+ "epoch": 0.5402087170042971,
+ "eval_entropy": 0.7768624458994184,
+ "eval_loss": 0.7051481604576111,
+ "eval_mean_token_accuracy": 0.7991346962111336,
+ "eval_num_tokens": 565606.0,
+ "eval_runtime": 57.7442,
+ "eval_samples_per_second": 24.176,
+ "eval_steps_per_second": 3.031,
+ "step": 220
+ },
+ {
+ "entropy": 0.7731568738818169,
+ "epoch": 0.5893186003683242,
+ "grad_norm": 0.553688645362854,
+ "learning_rate": 0.00019164429385730688,
+ "loss": 0.7003726959228516,
+ "mean_token_accuracy": 0.7977605648338795,
+ "num_tokens": 616528.0,
+ "step": 240
+ },
+ {
+ "epoch": 0.5893186003683242,
+ "eval_entropy": 0.7570131056649344,
+ "eval_loss": 0.7042225003242493,
+ "eval_mean_token_accuracy": 0.8000479991095407,
+ "eval_num_tokens": 616528.0,
+ "eval_runtime": 57.8713,
+ "eval_samples_per_second": 24.122,
+ "eval_steps_per_second": 3.024,
+ "step": 240
+ },
+ {
+ "entropy": 0.7343776792287826,
+ "epoch": 0.6384284837323512,
+ "grad_norm": 0.5576229095458984,
+ "learning_rate": 0.00018561406922024723,
+ "loss": 0.6735261440277099,
+ "mean_token_accuracy": 0.8059290319681167,
+ "num_tokens": 669440.0,
+ "step": 260
+ },
+ {
+ "epoch": 0.6384284837323512,
+ "eval_entropy": 0.7231997234480722,
+ "eval_loss": 0.6926313638687134,
+ "eval_mean_token_accuracy": 0.8031776939119611,
+ "eval_num_tokens": 669440.0,
+ "eval_runtime": 57.1692,
+ "eval_samples_per_second": 24.419,
+ "eval_steps_per_second": 3.061,
+ "step": 260
+ },
+ {
+ "entropy": 0.7326226674020291,
+ "epoch": 0.6875383670963782,
+ "grad_norm": 0.5962154865264893,
+ "learning_rate": 0.00017901186135690458,
+ "loss": 0.6677989482879638,
+ "mean_token_accuracy": 0.8022022873163224,
+ "num_tokens": 720146.0,
+ "step": 280
+ },
+ {
+ "epoch": 0.6875383670963782,
+ "eval_entropy": 0.7265675282478332,
+ "eval_loss": 0.6867054104804993,
+ "eval_mean_token_accuracy": 0.8021906747136798,
+ "eval_num_tokens": 720146.0,
+ "eval_runtime": 57.8067,
+ "eval_samples_per_second": 24.149,
+ "eval_steps_per_second": 3.027,
+ "step": 280
+ },
+ {
+ "entropy": 0.7409031428396702,
+ "epoch": 0.7366482504604052,
+ "grad_norm": 0.5716915726661682,
+ "learning_rate": 0.0001718860197237135,
+ "loss": 0.6692957878112793,
+ "mean_token_accuracy": 0.8036864519119262,
+ "num_tokens": 772624.0,
+ "step": 300
+ },
+ {
+ "epoch": 0.7366482504604052,
+ "eval_entropy": 0.7272419779641288,
+ "eval_loss": 0.6771560907363892,
+ "eval_mean_token_accuracy": 0.8062130059514727,
+ "eval_num_tokens": 772624.0,
+ "eval_runtime": 57.5942,
+ "eval_samples_per_second": 24.239,
+ "eval_steps_per_second": 3.039,
+ "step": 300
+ },
+ {
+ "entropy": 0.7243741095066071,
+ "epoch": 0.7857581338244322,
+ "grad_norm": 0.5360459089279175,
+ "learning_rate": 0.00016428872846524118,
+ "loss": 0.6628021240234375,
+ "mean_token_accuracy": 0.8050650551915168,
+ "num_tokens": 822969.0,
+ "step": 320
+ },
+ {
+ "epoch": 0.7857581338244322,
+ "eval_entropy": 0.7325100105149406,
+ "eval_loss": 0.6667791604995728,
+ "eval_mean_token_accuracy": 0.8082476602281843,
+ "eval_num_tokens": 822969.0,
+ "eval_runtime": 57.3807,
+ "eval_samples_per_second": 24.329,
+ "eval_steps_per_second": 3.05,
+ "step": 320
+ },
+ {
+ "entropy": 0.7213595665991306,
+ "epoch": 0.8348680171884592,
+ "grad_norm": 0.48658841848373413,
+ "learning_rate": 0.00015627562425793356,
+ "loss": 0.6494232654571533,
+ "mean_token_accuracy": 0.811632139235735,
+ "num_tokens": 873459.0,
+ "step": 340
+ },
+ {
+ "epoch": 0.8348680171884592,
+ "eval_entropy": 0.7198746153286525,
+ "eval_loss": 0.6634471416473389,
+ "eval_mean_token_accuracy": 0.8094694914136614,
+ "eval_num_tokens": 873459.0,
+ "eval_runtime": 57.6988,
+ "eval_samples_per_second": 24.195,
+ "eval_steps_per_second": 3.033,
+ "step": 340
+ },
+ {
+ "entropy": 0.7133754774928093,
+ "epoch": 0.8839779005524862,
+ "grad_norm": 0.685793399810791,
+ "learning_rate": 0.00014790538887019238,
+ "loss": 0.642530870437622,
+ "mean_token_accuracy": 0.8122274152934551,
+ "num_tokens": 925669.0,
+ "step": 360
+ },
+ {
+ "epoch": 0.8839779005524862,
+ "eval_entropy": 0.6683637884684971,
+ "eval_loss": 0.6579923033714294,
+ "eval_mean_token_accuracy": 0.811241352558136,
+ "eval_num_tokens": 925669.0,
+ "eval_runtime": 57.8503,
+ "eval_samples_per_second": 24.131,
+ "eval_steps_per_second": 3.025,
+ "step": 360
+ },
+ {
+ "entropy": 0.7262612670660019,
+ "epoch": 0.9330877839165131,
+ "grad_norm": 0.6348966956138611,
+ "learning_rate": 0.00013923931942255727,
+ "loss": 0.656022596359253,
+ "mean_token_accuracy": 0.8072879828512669,
+ "num_tokens": 973436.0,
+ "step": 380
+ },
+ {
+ "epoch": 0.9330877839165131,
+ "eval_entropy": 0.7012172797748021,
+ "eval_loss": 0.651187002658844,
+ "eval_mean_token_accuracy": 0.8117410482679095,
+ "eval_num_tokens": 973436.0,
+ "eval_runtime": 57.1997,
+ "eval_samples_per_second": 24.406,
+ "eval_steps_per_second": 3.059,
+ "step": 380
+ },
+ {
+ "entropy": 0.7007582865655422,
+ "epoch": 0.9821976672805403,
+ "grad_norm": 0.4913424849510193,
+ "learning_rate": 0.0001303408794950743,
+ "loss": 0.6367915153503418,
+ "mean_token_accuracy": 0.8122764162719249,
+ "num_tokens": 1026370.0,
+ "step": 400
+ },
+ {
+ "epoch": 0.9821976672805403,
+ "eval_entropy": 0.6719393570082528,
+ "eval_loss": 0.6448967456817627,
+ "eval_mean_token_accuracy": 0.8140764743941171,
+ "eval_num_tokens": 1026370.0,
+ "eval_runtime": 57.9405,
+ "eval_samples_per_second": 24.094,
+ "eval_steps_per_second": 3.02,
+ "step": 400
+ },
+ {
+ "entropy": 0.6647140879135627,
+ "epoch": 1.0294659300184161,
+ "grad_norm": 0.4148324429988861,
+ "learning_rate": 0.00012127523436919402,
+ "loss": 0.5996760845184326,
+ "mean_token_accuracy": 0.8205991890523341,
+ "num_tokens": 1076273.0,
+ "step": 420
+ },
+ {
+ "epoch": 1.0294659300184161,
+ "eval_entropy": 0.6531744088445391,
+ "eval_loss": 0.6474375128746033,
+ "eval_mean_token_accuracy": 0.8118055742127555,
+ "eval_num_tokens": 1076273.0,
+ "eval_runtime": 57.7161,
+ "eval_samples_per_second": 24.187,
+ "eval_steps_per_second": 3.032,
+ "step": 420
+ },
+ {
+ "entropy": 0.64856808334589,
+ "epoch": 1.0785758133824432,
+ "grad_norm": 0.4203054904937744,
+ "learning_rate": 0.0001121087738077257,
+ "loss": 0.580596923828125,
+ "mean_token_accuracy": 0.8229757599532604,
+ "num_tokens": 1133367.0,
+ "step": 440
+ },
+ {
+ "epoch": 1.0785758133824432,
+ "eval_entropy": 0.6491447843824114,
+ "eval_loss": 0.6415509581565857,
+ "eval_mean_token_accuracy": 0.8132358625956944,
+ "eval_num_tokens": 1133367.0,
+ "eval_runtime": 57.0688,
+ "eval_samples_per_second": 24.462,
+ "eval_steps_per_second": 3.066,
+ "step": 440
+ },
+ {
+ "entropy": 0.6387462258338928,
+ "epoch": 1.1276856967464703,
+ "grad_norm": 0.4547317326068878,
+ "learning_rate": 0.0001029086258676389,
+ "loss": 0.5705233097076416,
+ "mean_token_accuracy": 0.8258621700108051,
+ "num_tokens": 1186585.0,
+ "step": 460
+ },
+ {
+ "epoch": 1.1276856967464703,
+ "eval_entropy": 0.653609961441585,
+ "eval_loss": 0.6332530975341797,
+ "eval_mean_token_accuracy": 0.8161680047852652,
+ "eval_num_tokens": 1186585.0,
+ "eval_runtime": 57.3472,
+ "eval_samples_per_second": 24.343,
+ "eval_steps_per_second": 3.052,
+ "step": 460
+ },
+ {
+ "entropy": 0.6427005555480718,
+ "epoch": 1.1767955801104972,
+ "grad_norm": 0.4293302297592163,
+ "learning_rate": 9.374216530617052e-05,
+ "loss": 0.5639504432678223,
+ "mean_token_accuracy": 0.8271843157708645,
+ "num_tokens": 1235457.0,
+ "step": 480
+ },
+ {
+ "epoch": 1.1767955801104972,
+ "eval_entropy": 0.6493809008598328,
+ "eval_loss": 0.6308221817016602,
+ "eval_mean_token_accuracy": 0.8174078563281468,
+ "eval_num_tokens": 1235457.0,
+ "eval_runtime": 57.7158,
+ "eval_samples_per_second": 24.187,
+ "eval_steps_per_second": 3.032,
+ "step": 480
+ },
+ {
+ "entropy": 0.617260966822505,
+ "epoch": 1.2259054634745243,
+ "grad_norm": 0.44997501373291016,
+ "learning_rate": 8.46765201802903e-05,
+ "loss": 0.5545801639556884,
+ "mean_token_accuracy": 0.8310177803039551,
+ "num_tokens": 1289995.0,
+ "step": 500
+ },
+ {
+ "epoch": 1.2259054634745243,
+ "eval_entropy": 0.6463884047099522,
+ "eval_loss": 0.6248639822006226,
+ "eval_mean_token_accuracy": 0.8196247465269906,
+ "eval_num_tokens": 1289995.0,
+ "eval_runtime": 57.655,
+ "eval_samples_per_second": 24.213,
+ "eval_steps_per_second": 3.035,
+ "step": 500
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 2.162562792483264e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..49f099052acaaa62e5a9a6914ef6590be09f26d9
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/trainer_state.json
@@ -0,0 +1,97 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.14732965009208104,
+ "eval_steps": 20,
+ "global_step": 60,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 2661782383219200.0,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231547eca790e866e907eade83a12f382230c57
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 128,
+ "lora_bias": false,
+ "lora_dropout": 0.056708081859357024,
+ "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": 128,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ccc6fc777cca7152d025ec97b396eae99ecce8e2
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/trainer_state.json
@@ -0,0 +1,118 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.19643953345610804,
+ "eval_steps": 20,
+ "global_step": 80,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.136037287116051,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 1.5031570196151733,
+ "learning_rate": 4.982110480282838e-05,
+ "loss": 1.9729637145996093,
+ "mean_token_accuracy": 0.5955793052911759,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 1.2274160248892647,
+ "eval_loss": 1.211025595664978,
+ "eval_mean_token_accuracy": 0.6963188947950091,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 58.8096,
+ "eval_samples_per_second": 23.738,
+ "eval_steps_per_second": 2.976,
+ "step": 20
+ },
+ {
+ "entropy": 1.0958118669688701,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.0549366474151611,
+ "learning_rate": 0.00010226437301633193,
+ "loss": 1.0308157920837402,
+ "mean_token_accuracy": 0.7261112071573734,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 0.9666143652370998,
+ "eval_loss": 0.9403602480888367,
+ "eval_mean_token_accuracy": 0.7466924602644784,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 57.3655,
+ "eval_samples_per_second": 24.335,
+ "eval_steps_per_second": 3.051,
+ "step": 40
+ },
+ {
+ "entropy": 0.9621845617890358,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 0.8900883793830872,
+ "learning_rate": 0.00015470764122983547,
+ "loss": 0.8886768341064453,
+ "mean_token_accuracy": 0.7551977805793285,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 0.9346491888591222,
+ "eval_loss": 0.8509624600410461,
+ "eval_mean_token_accuracy": 0.7676705942835127,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 57.519,
+ "eval_samples_per_second": 24.27,
+ "eval_steps_per_second": 3.042,
+ "step": 60
+ },
+ {
+ "entropy": 0.9030815377831459,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 0.9833227396011353,
+ "learning_rate": 0.00020715090944333904,
+ "loss": 0.8231673240661621,
+ "mean_token_accuracy": 0.7664932057261467,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.8813995848383223,
+ "eval_loss": 0.8210445642471313,
+ "eval_mean_token_accuracy": 0.7710604834556579,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 57.755,
+ "eval_samples_per_second": 24.171,
+ "eval_steps_per_second": 3.03,
+ "step": 80
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 3555048999029760.0,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..9b9a4dda9e8e58ad1514a0a6387528f54a2c30c5
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/trainer_state.json
@@ -0,0 +1,538 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 1.1767955801104972,
+ "eval_steps": 20,
+ "global_step": 480,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.5655054792761804,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 2.106731653213501,
+ "learning_rate": 1.9428070010160535e-05,
+ "loss": 2.3814184188842775,
+ "mean_token_accuracy": 0.5490218658000231,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 2.196743746485029,
+ "eval_loss": 2.028722047805786,
+ "eval_mean_token_accuracy": 0.5899124583176204,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 55.8819,
+ "eval_samples_per_second": 24.963,
+ "eval_steps_per_second": 3.132,
+ "step": 20
+ },
+ {
+ "entropy": 1.6433217093348502,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.7636586427688599,
+ "learning_rate": 3.987867002085583e-05,
+ "loss": 1.4883951187133788,
+ "mean_token_accuracy": 0.6520100250840187,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 1.2071456554957798,
+ "eval_loss": 1.1546095609664917,
+ "eval_mean_token_accuracy": 0.7017311205182757,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 55.21,
+ "eval_samples_per_second": 25.267,
+ "eval_steps_per_second": 3.17,
+ "step": 40
+ },
+ {
+ "entropy": 1.1379728086292744,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 1.6891512870788574,
+ "learning_rate": 6.032927003155113e-05,
+ "loss": 1.0574092864990234,
+ "mean_token_accuracy": 0.7190469801425934,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 1.026683736869267,
+ "eval_loss": 0.9847686290740967,
+ "eval_mean_token_accuracy": 0.7383982699257987,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 54.8863,
+ "eval_samples_per_second": 25.416,
+ "eval_steps_per_second": 3.188,
+ "step": 60
+ },
+ {
+ "entropy": 1.0109580919146537,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 1.7396622896194458,
+ "learning_rate": 8.077987004224643e-05,
+ "loss": 0.9192052841186523,
+ "mean_token_accuracy": 0.7471949681639671,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.9901534714017596,
+ "eval_loss": 0.8948126435279846,
+ "eval_mean_token_accuracy": 0.7570484624590192,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 54.924,
+ "eval_samples_per_second": 25.399,
+ "eval_steps_per_second": 3.186,
+ "step": 80
+ },
+ {
+ "entropy": 0.9607135757803917,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 2.213541030883789,
+ "learning_rate": 8.373653138200974e-05,
+ "loss": 0.8742897987365723,
+ "mean_token_accuracy": 0.7584197156131267,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.9265402409008571,
+ "eval_loss": 0.8480852246284485,
+ "eval_mean_token_accuracy": 0.7668960053580148,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 54.5404,
+ "eval_samples_per_second": 25.577,
+ "eval_steps_per_second": 3.209,
+ "step": 100
+ },
+ {
+ "entropy": 0.9136901050806046,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 1.2657700777053833,
+ "learning_rate": 8.332285415330192e-05,
+ "loss": 0.8283636093139648,
+ "mean_token_accuracy": 0.7690222583711147,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8823672638620649,
+ "eval_loss": 0.8114516139030457,
+ "eval_mean_token_accuracy": 0.7755525919369289,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 55.3013,
+ "eval_samples_per_second": 25.225,
+ "eval_steps_per_second": 3.164,
+ "step": 120
+ },
+ {
+ "entropy": 0.8614003121852875,
+ "epoch": 0.3437691835481891,
+ "grad_norm": 1.0693161487579346,
+ "learning_rate": 8.260600180650497e-05,
+ "loss": 0.778686809539795,
+ "mean_token_accuracy": 0.7784358590841294,
+ "num_tokens": 361815.0,
+ "step": 140
+ },
+ {
+ "epoch": 0.3437691835481891,
+ "eval_entropy": 0.8277448885781424,
+ "eval_loss": 0.7953144311904907,
+ "eval_mean_token_accuracy": 0.7796006223133632,
+ "eval_num_tokens": 361815.0,
+ "eval_runtime": 55.4206,
+ "eval_samples_per_second": 25.171,
+ "eval_steps_per_second": 3.158,
+ "step": 140
+ },
+ {
+ "entropy": 0.8520236104726792,
+ "epoch": 0.3928790669122161,
+ "grad_norm": 1.038674235343933,
+ "learning_rate": 8.159122401294533e-05,
+ "loss": 0.7658556938171387,
+ "mean_token_accuracy": 0.7823643557727337,
+ "num_tokens": 417429.0,
+ "step": 160
+ },
+ {
+ "epoch": 0.3928790669122161,
+ "eval_entropy": 0.8204359361103603,
+ "eval_loss": 0.7691255807876587,
+ "eval_mean_token_accuracy": 0.7845319850104195,
+ "eval_num_tokens": 417429.0,
+ "eval_runtime": 55.1624,
+ "eval_samples_per_second": 25.289,
+ "eval_steps_per_second": 3.172,
+ "step": 160
+ },
+ {
+ "entropy": 0.8366772942245007,
+ "epoch": 0.4419889502762431,
+ "grad_norm": 1.441671371459961,
+ "learning_rate": 8.028595221915362e-05,
+ "loss": 0.7603907108306884,
+ "mean_token_accuracy": 0.7837655946612359,
+ "num_tokens": 467817.0,
+ "step": 180
+ },
+ {
+ "epoch": 0.4419889502762431,
+ "eval_entropy": 0.8421429433141436,
+ "eval_loss": 0.7569069862365723,
+ "eval_mean_token_accuracy": 0.7865874515260969,
+ "eval_num_tokens": 467817.0,
+ "eval_runtime": 55.1833,
+ "eval_samples_per_second": 25.279,
+ "eval_steps_per_second": 3.171,
+ "step": 180
+ },
+ {
+ "entropy": 0.8282789587974548,
+ "epoch": 0.4910988336402701,
+ "grad_norm": 1.1421457529067993,
+ "learning_rate": 7.86997452247065e-05,
+ "loss": 0.7440836906433106,
+ "mean_token_accuracy": 0.7888234727084636,
+ "num_tokens": 513773.0,
+ "step": 200
+ },
+ {
+ "epoch": 0.4910988336402701,
+ "eval_entropy": 0.8229101082256862,
+ "eval_loss": 0.7469918131828308,
+ "eval_mean_token_accuracy": 0.7869657233783177,
+ "eval_num_tokens": 513773.0,
+ "eval_runtime": 54.7451,
+ "eval_samples_per_second": 25.482,
+ "eval_steps_per_second": 3.197,
+ "step": 200
+ },
+ {
+ "entropy": 0.8064627222716808,
+ "epoch": 0.5402087170042971,
+ "grad_norm": 1.1229082345962524,
+ "learning_rate": 7.684421918098267e-05,
+ "loss": 0.732261323928833,
+ "mean_token_accuracy": 0.7928754203021526,
+ "num_tokens": 565606.0,
+ "step": 220
+ },
+ {
+ "epoch": 0.5402087170042971,
+ "eval_entropy": 0.8189267148290361,
+ "eval_loss": 0.7362269759178162,
+ "eval_mean_token_accuracy": 0.7914395029204232,
+ "eval_num_tokens": 565606.0,
+ "eval_runtime": 54.9856,
+ "eval_samples_per_second": 25.37,
+ "eval_steps_per_second": 3.183,
+ "step": 220
+ },
+ {
+ "entropy": 0.8221734672784805,
+ "epoch": 0.5893186003683242,
+ "grad_norm": 1.1489790678024292,
+ "learning_rate": 7.473296252346783e-05,
+ "loss": 0.7320152282714844,
+ "mean_token_accuracy": 0.789833651483059,
+ "num_tokens": 616528.0,
+ "step": 240
+ },
+ {
+ "epoch": 0.5893186003683242,
+ "eval_entropy": 0.8049671122006008,
+ "eval_loss": 0.728659987449646,
+ "eval_mean_token_accuracy": 0.7936891293525696,
+ "eval_num_tokens": 616528.0,
+ "eval_runtime": 54.285,
+ "eval_samples_per_second": 25.698,
+ "eval_steps_per_second": 3.224,
+ "step": 240
+ },
+ {
+ "entropy": 0.7805217795073986,
+ "epoch": 0.6384284837323512,
+ "grad_norm": 0.9496471285820007,
+ "learning_rate": 7.238143646057855e-05,
+ "loss": 0.7050290584564209,
+ "mean_token_accuracy": 0.7986594937741757,
+ "num_tokens": 669440.0,
+ "step": 260
+ },
+ {
+ "epoch": 0.6384284837323512,
+ "eval_entropy": 0.7794270893505642,
+ "eval_loss": 0.7148454785346985,
+ "eval_mean_token_accuracy": 0.7973272456441607,
+ "eval_num_tokens": 669440.0,
+ "eval_runtime": 53.9884,
+ "eval_samples_per_second": 25.839,
+ "eval_steps_per_second": 3.241,
+ "step": 260
+ },
+ {
+ "entropy": 0.781412948668003,
+ "epoch": 0.6875383670963782,
+ "grad_norm": 1.1252366304397583,
+ "learning_rate": 6.980686174774777e-05,
+ "loss": 0.6978650093078613,
+ "mean_token_accuracy": 0.7952493160963059,
+ "num_tokens": 720146.0,
+ "step": 280
+ },
+ {
+ "epoch": 0.6875383670963782,
+ "eval_entropy": 0.7790967045511519,
+ "eval_loss": 0.7110744118690491,
+ "eval_mean_token_accuracy": 0.7964006056104388,
+ "eval_num_tokens": 720146.0,
+ "eval_runtime": 53.9404,
+ "eval_samples_per_second": 25.862,
+ "eval_steps_per_second": 3.244,
+ "step": 280
+ },
+ {
+ "entropy": 0.7788082852959632,
+ "epoch": 0.7366482504604052,
+ "grad_norm": 1.1519086360931396,
+ "learning_rate": 6.702809257595105e-05,
+ "loss": 0.7019347190856934,
+ "mean_token_accuracy": 0.7960198536515236,
+ "num_tokens": 772624.0,
+ "step": 300
+ },
+ {
+ "epoch": 0.7366482504604052,
+ "eval_entropy": 0.7932151372092111,
+ "eval_loss": 0.7035816311836243,
+ "eval_mean_token_accuracy": 0.7979611914498466,
+ "eval_num_tokens": 772624.0,
+ "eval_runtime": 54.0517,
+ "eval_samples_per_second": 25.809,
+ "eval_steps_per_second": 3.238,
+ "step": 300
+ },
+ {
+ "entropy": 0.7729029789566993,
+ "epoch": 0.7857581338244322,
+ "grad_norm": 1.0247046947479248,
+ "learning_rate": 6.406547849821584e-05,
+ "loss": 0.6924194812774658,
+ "mean_token_accuracy": 0.7985413707792759,
+ "num_tokens": 822969.0,
+ "step": 320
+ },
+ {
+ "epoch": 0.7857581338244322,
+ "eval_entropy": 0.7812128639221192,
+ "eval_loss": 0.6989337205886841,
+ "eval_mean_token_accuracy": 0.8005841762678964,
+ "eval_num_tokens": 822969.0,
+ "eval_runtime": 54.1306,
+ "eval_samples_per_second": 25.771,
+ "eval_steps_per_second": 3.233,
+ "step": 320
+ },
+ {
+ "entropy": 0.7627703189849854,
+ "epoch": 0.8348680171884592,
+ "grad_norm": 0.957506537437439,
+ "learning_rate": 6.0940715405257566e-05,
+ "loss": 0.679268217086792,
+ "mean_token_accuracy": 0.8040587857365609,
+ "num_tokens": 873459.0,
+ "step": 340
+ },
+ {
+ "epoch": 0.8348680171884592,
+ "eval_entropy": 0.7624005011149815,
+ "eval_loss": 0.6934819221496582,
+ "eval_mean_token_accuracy": 0.8014607231957572,
+ "eval_num_tokens": 873459.0,
+ "eval_runtime": 54.1656,
+ "eval_samples_per_second": 25.754,
+ "eval_steps_per_second": 3.231,
+ "step": 340
+ },
+ {
+ "entropy": 0.763699171692133,
+ "epoch": 0.8839779005524862,
+ "grad_norm": 1.4329766035079956,
+ "learning_rate": 5.767668664158135e-05,
+ "loss": 0.6764235019683837,
+ "mean_token_accuracy": 0.8051542706787587,
+ "num_tokens": 925669.0,
+ "step": 360
+ },
+ {
+ "epoch": 0.8839779005524862,
+ "eval_entropy": 0.7163350166593279,
+ "eval_loss": 0.6882685422897339,
+ "eval_mean_token_accuracy": 0.802961312021528,
+ "eval_num_tokens": 925669.0,
+ "eval_runtime": 53.8288,
+ "eval_samples_per_second": 25.916,
+ "eval_steps_per_second": 3.251,
+ "step": 360
+ },
+ {
+ "entropy": 0.7710359543561935,
+ "epoch": 0.9330877839165131,
+ "grad_norm": 1.141253113746643,
+ "learning_rate": 5.429729542559189e-05,
+ "loss": 0.6897628784179688,
+ "mean_token_accuracy": 0.8005676351487636,
+ "num_tokens": 973436.0,
+ "step": 380
+ },
+ {
+ "epoch": 0.9330877839165131,
+ "eval_entropy": 0.7462973778588431,
+ "eval_loss": 0.6829195022583008,
+ "eval_mean_token_accuracy": 0.8036447630609784,
+ "eval_num_tokens": 973436.0,
+ "eval_runtime": 53.8276,
+ "eval_samples_per_second": 25.916,
+ "eval_steps_per_second": 3.251,
+ "step": 380
+ },
+ {
+ "entropy": 0.7500453963875771,
+ "epoch": 0.9821976672805403,
+ "grad_norm": 0.9055600166320801,
+ "learning_rate": 5.0827289800936774e-05,
+ "loss": 0.6726306915283203,
+ "mean_token_accuracy": 0.8026990614831447,
+ "num_tokens": 1026370.0,
+ "step": 400
+ },
+ {
+ "epoch": 0.9821976672805403,
+ "eval_entropy": 0.745090776511601,
+ "eval_loss": 0.6795706152915955,
+ "eval_mean_token_accuracy": 0.8038556255613054,
+ "eval_num_tokens": 1026370.0,
+ "eval_runtime": 53.9485,
+ "eval_samples_per_second": 25.858,
+ "eval_steps_per_second": 3.244,
+ "step": 400
+ },
+ {
+ "entropy": 0.7288912821125675,
+ "epoch": 1.0294659300184161,
+ "grad_norm": 0.9075890779495239,
+ "learning_rate": 4.729208140100435e-05,
+ "loss": 0.6533905506134033,
+ "mean_token_accuracy": 0.807745551908171,
+ "num_tokens": 1076273.0,
+ "step": 420
+ },
+ {
+ "epoch": 1.0294659300184161,
+ "eval_entropy": 0.7279435137340001,
+ "eval_loss": 0.6758045554161072,
+ "eval_mean_token_accuracy": 0.8043134975433349,
+ "eval_num_tokens": 1076273.0,
+ "eval_runtime": 53.8173,
+ "eval_samples_per_second": 25.921,
+ "eval_steps_per_second": 3.252,
+ "step": 420
+ },
+ {
+ "entropy": 0.7183651916682721,
+ "epoch": 1.0785758133824432,
+ "grad_norm": 0.9271263480186462,
+ "learning_rate": 4.371755935380415e-05,
+ "loss": 0.6382484912872315,
+ "mean_token_accuracy": 0.810582410544157,
+ "num_tokens": 1133367.0,
+ "step": 440
+ },
+ {
+ "epoch": 1.0785758133824432,
+ "eval_entropy": 0.7224361307280405,
+ "eval_loss": 0.6700165271759033,
+ "eval_mean_token_accuracy": 0.8052176751409258,
+ "eval_num_tokens": 1133367.0,
+ "eval_runtime": 54.2346,
+ "eval_samples_per_second": 25.722,
+ "eval_steps_per_second": 3.227,
+ "step": 440
+ },
+ {
+ "entropy": 0.7097452029585838,
+ "epoch": 1.1276856967464703,
+ "grad_norm": 0.8726329207420349,
+ "learning_rate": 4.012990069004659e-05,
+ "loss": 0.6309147834777832,
+ "mean_token_accuracy": 0.8141021691262722,
+ "num_tokens": 1186585.0,
+ "step": 460
+ },
+ {
+ "epoch": 1.1276856967464703,
+ "eval_entropy": 0.7231041298593793,
+ "eval_loss": 0.6672903299331665,
+ "eval_mean_token_accuracy": 0.8079326885087149,
+ "eval_num_tokens": 1186585.0,
+ "eval_runtime": 53.7923,
+ "eval_samples_per_second": 25.933,
+ "eval_steps_per_second": 3.253,
+ "step": 460
+ },
+ {
+ "entropy": 0.7170382678508759,
+ "epoch": 1.1767955801104972,
+ "grad_norm": 0.8970508575439453,
+ "learning_rate": 3.6555378642846364e-05,
+ "loss": 0.6294277667999267,
+ "mean_token_accuracy": 0.8117733597755432,
+ "num_tokens": 1235457.0,
+ "step": 480
+ },
+ {
+ "epoch": 1.1767955801104972,
+ "eval_entropy": 0.7064940530913216,
+ "eval_loss": 0.6625319123268127,
+ "eval_mean_token_accuracy": 0.8093922696794782,
+ "eval_num_tokens": 1235457.0,
+ "eval_runtime": 53.6574,
+ "eval_samples_per_second": 25.998,
+ "eval_steps_per_second": 3.261,
+ "step": 480
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.980221720242368e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6d05047585cef4bba4ea582a1f394fe467801e35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 64,
+ "lora_bias": false,
+ "lora_dropout": 0.055708383706168485,
+ "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": 32,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c6f3eac6c0aaf9262862579223f8aa385cf431fe
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/trainer_state.json
@@ -0,0 +1,559 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 1.2259054634745243,
+ "eval_steps": 20,
+ "global_step": 500,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.5655054792761804,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 2.106731653213501,
+ "learning_rate": 1.9428070010160535e-05,
+ "loss": 2.3814184188842775,
+ "mean_token_accuracy": 0.5490218658000231,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 2.196743746485029,
+ "eval_loss": 2.028722047805786,
+ "eval_mean_token_accuracy": 0.5899124583176204,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 55.8819,
+ "eval_samples_per_second": 24.963,
+ "eval_steps_per_second": 3.132,
+ "step": 20
+ },
+ {
+ "entropy": 1.6433217093348502,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.7636586427688599,
+ "learning_rate": 3.987867002085583e-05,
+ "loss": 1.4883951187133788,
+ "mean_token_accuracy": 0.6520100250840187,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 1.2071456554957798,
+ "eval_loss": 1.1546095609664917,
+ "eval_mean_token_accuracy": 0.7017311205182757,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 55.21,
+ "eval_samples_per_second": 25.267,
+ "eval_steps_per_second": 3.17,
+ "step": 40
+ },
+ {
+ "entropy": 1.1379728086292744,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 1.6891512870788574,
+ "learning_rate": 6.032927003155113e-05,
+ "loss": 1.0574092864990234,
+ "mean_token_accuracy": 0.7190469801425934,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 1.026683736869267,
+ "eval_loss": 0.9847686290740967,
+ "eval_mean_token_accuracy": 0.7383982699257987,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 54.8863,
+ "eval_samples_per_second": 25.416,
+ "eval_steps_per_second": 3.188,
+ "step": 60
+ },
+ {
+ "entropy": 1.0109580919146537,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 1.7396622896194458,
+ "learning_rate": 8.077987004224643e-05,
+ "loss": 0.9192052841186523,
+ "mean_token_accuracy": 0.7471949681639671,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.9901534714017596,
+ "eval_loss": 0.8948126435279846,
+ "eval_mean_token_accuracy": 0.7570484624590192,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 54.924,
+ "eval_samples_per_second": 25.399,
+ "eval_steps_per_second": 3.186,
+ "step": 80
+ },
+ {
+ "entropy": 0.9607135757803917,
+ "epoch": 0.24554941682013506,
+ "grad_norm": 2.213541030883789,
+ "learning_rate": 8.373653138200974e-05,
+ "loss": 0.8742897987365723,
+ "mean_token_accuracy": 0.7584197156131267,
+ "num_tokens": 253606.0,
+ "step": 100
+ },
+ {
+ "epoch": 0.24554941682013506,
+ "eval_entropy": 0.9265402409008571,
+ "eval_loss": 0.8480852246284485,
+ "eval_mean_token_accuracy": 0.7668960053580148,
+ "eval_num_tokens": 253606.0,
+ "eval_runtime": 54.5404,
+ "eval_samples_per_second": 25.577,
+ "eval_steps_per_second": 3.209,
+ "step": 100
+ },
+ {
+ "entropy": 0.9136901050806046,
+ "epoch": 0.2946593001841621,
+ "grad_norm": 1.2657700777053833,
+ "learning_rate": 8.332285415330192e-05,
+ "loss": 0.8283636093139648,
+ "mean_token_accuracy": 0.7690222583711147,
+ "num_tokens": 306583.0,
+ "step": 120
+ },
+ {
+ "epoch": 0.2946593001841621,
+ "eval_entropy": 0.8823672638620649,
+ "eval_loss": 0.8114516139030457,
+ "eval_mean_token_accuracy": 0.7755525919369289,
+ "eval_num_tokens": 306583.0,
+ "eval_runtime": 55.3013,
+ "eval_samples_per_second": 25.225,
+ "eval_steps_per_second": 3.164,
+ "step": 120
+ },
+ {
+ "entropy": 0.8614003121852875,
+ "epoch": 0.3437691835481891,
+ "grad_norm": 1.0693161487579346,
+ "learning_rate": 8.260600180650497e-05,
+ "loss": 0.778686809539795,
+ "mean_token_accuracy": 0.7784358590841294,
+ "num_tokens": 361815.0,
+ "step": 140
+ },
+ {
+ "epoch": 0.3437691835481891,
+ "eval_entropy": 0.8277448885781424,
+ "eval_loss": 0.7953144311904907,
+ "eval_mean_token_accuracy": 0.7796006223133632,
+ "eval_num_tokens": 361815.0,
+ "eval_runtime": 55.4206,
+ "eval_samples_per_second": 25.171,
+ "eval_steps_per_second": 3.158,
+ "step": 140
+ },
+ {
+ "entropy": 0.8520236104726792,
+ "epoch": 0.3928790669122161,
+ "grad_norm": 1.038674235343933,
+ "learning_rate": 8.159122401294533e-05,
+ "loss": 0.7658556938171387,
+ "mean_token_accuracy": 0.7823643557727337,
+ "num_tokens": 417429.0,
+ "step": 160
+ },
+ {
+ "epoch": 0.3928790669122161,
+ "eval_entropy": 0.8204359361103603,
+ "eval_loss": 0.7691255807876587,
+ "eval_mean_token_accuracy": 0.7845319850104195,
+ "eval_num_tokens": 417429.0,
+ "eval_runtime": 55.1624,
+ "eval_samples_per_second": 25.289,
+ "eval_steps_per_second": 3.172,
+ "step": 160
+ },
+ {
+ "entropy": 0.8366772942245007,
+ "epoch": 0.4419889502762431,
+ "grad_norm": 1.441671371459961,
+ "learning_rate": 8.028595221915362e-05,
+ "loss": 0.7603907108306884,
+ "mean_token_accuracy": 0.7837655946612359,
+ "num_tokens": 467817.0,
+ "step": 180
+ },
+ {
+ "epoch": 0.4419889502762431,
+ "eval_entropy": 0.8421429433141436,
+ "eval_loss": 0.7569069862365723,
+ "eval_mean_token_accuracy": 0.7865874515260969,
+ "eval_num_tokens": 467817.0,
+ "eval_runtime": 55.1833,
+ "eval_samples_per_second": 25.279,
+ "eval_steps_per_second": 3.171,
+ "step": 180
+ },
+ {
+ "entropy": 0.8282789587974548,
+ "epoch": 0.4910988336402701,
+ "grad_norm": 1.1421457529067993,
+ "learning_rate": 7.86997452247065e-05,
+ "loss": 0.7440836906433106,
+ "mean_token_accuracy": 0.7888234727084636,
+ "num_tokens": 513773.0,
+ "step": 200
+ },
+ {
+ "epoch": 0.4910988336402701,
+ "eval_entropy": 0.8229101082256862,
+ "eval_loss": 0.7469918131828308,
+ "eval_mean_token_accuracy": 0.7869657233783177,
+ "eval_num_tokens": 513773.0,
+ "eval_runtime": 54.7451,
+ "eval_samples_per_second": 25.482,
+ "eval_steps_per_second": 3.197,
+ "step": 200
+ },
+ {
+ "entropy": 0.8064627222716808,
+ "epoch": 0.5402087170042971,
+ "grad_norm": 1.1229082345962524,
+ "learning_rate": 7.684421918098267e-05,
+ "loss": 0.732261323928833,
+ "mean_token_accuracy": 0.7928754203021526,
+ "num_tokens": 565606.0,
+ "step": 220
+ },
+ {
+ "epoch": 0.5402087170042971,
+ "eval_entropy": 0.8189267148290361,
+ "eval_loss": 0.7362269759178162,
+ "eval_mean_token_accuracy": 0.7914395029204232,
+ "eval_num_tokens": 565606.0,
+ "eval_runtime": 54.9856,
+ "eval_samples_per_second": 25.37,
+ "eval_steps_per_second": 3.183,
+ "step": 220
+ },
+ {
+ "entropy": 0.8221734672784805,
+ "epoch": 0.5893186003683242,
+ "grad_norm": 1.1489790678024292,
+ "learning_rate": 7.473296252346783e-05,
+ "loss": 0.7320152282714844,
+ "mean_token_accuracy": 0.789833651483059,
+ "num_tokens": 616528.0,
+ "step": 240
+ },
+ {
+ "epoch": 0.5893186003683242,
+ "eval_entropy": 0.8049671122006008,
+ "eval_loss": 0.728659987449646,
+ "eval_mean_token_accuracy": 0.7936891293525696,
+ "eval_num_tokens": 616528.0,
+ "eval_runtime": 54.285,
+ "eval_samples_per_second": 25.698,
+ "eval_steps_per_second": 3.224,
+ "step": 240
+ },
+ {
+ "entropy": 0.7805217795073986,
+ "epoch": 0.6384284837323512,
+ "grad_norm": 0.9496471285820007,
+ "learning_rate": 7.238143646057855e-05,
+ "loss": 0.7050290584564209,
+ "mean_token_accuracy": 0.7986594937741757,
+ "num_tokens": 669440.0,
+ "step": 260
+ },
+ {
+ "epoch": 0.6384284837323512,
+ "eval_entropy": 0.7794270893505642,
+ "eval_loss": 0.7148454785346985,
+ "eval_mean_token_accuracy": 0.7973272456441607,
+ "eval_num_tokens": 669440.0,
+ "eval_runtime": 53.9884,
+ "eval_samples_per_second": 25.839,
+ "eval_steps_per_second": 3.241,
+ "step": 260
+ },
+ {
+ "entropy": 0.781412948668003,
+ "epoch": 0.6875383670963782,
+ "grad_norm": 1.1252366304397583,
+ "learning_rate": 6.980686174774777e-05,
+ "loss": 0.6978650093078613,
+ "mean_token_accuracy": 0.7952493160963059,
+ "num_tokens": 720146.0,
+ "step": 280
+ },
+ {
+ "epoch": 0.6875383670963782,
+ "eval_entropy": 0.7790967045511519,
+ "eval_loss": 0.7110744118690491,
+ "eval_mean_token_accuracy": 0.7964006056104388,
+ "eval_num_tokens": 720146.0,
+ "eval_runtime": 53.9404,
+ "eval_samples_per_second": 25.862,
+ "eval_steps_per_second": 3.244,
+ "step": 280
+ },
+ {
+ "entropy": 0.7788082852959632,
+ "epoch": 0.7366482504604052,
+ "grad_norm": 1.1519086360931396,
+ "learning_rate": 6.702809257595105e-05,
+ "loss": 0.7019347190856934,
+ "mean_token_accuracy": 0.7960198536515236,
+ "num_tokens": 772624.0,
+ "step": 300
+ },
+ {
+ "epoch": 0.7366482504604052,
+ "eval_entropy": 0.7932151372092111,
+ "eval_loss": 0.7035816311836243,
+ "eval_mean_token_accuracy": 0.7979611914498466,
+ "eval_num_tokens": 772624.0,
+ "eval_runtime": 54.0517,
+ "eval_samples_per_second": 25.809,
+ "eval_steps_per_second": 3.238,
+ "step": 300
+ },
+ {
+ "entropy": 0.7729029789566993,
+ "epoch": 0.7857581338244322,
+ "grad_norm": 1.0247046947479248,
+ "learning_rate": 6.406547849821584e-05,
+ "loss": 0.6924194812774658,
+ "mean_token_accuracy": 0.7985413707792759,
+ "num_tokens": 822969.0,
+ "step": 320
+ },
+ {
+ "epoch": 0.7857581338244322,
+ "eval_entropy": 0.7812128639221192,
+ "eval_loss": 0.6989337205886841,
+ "eval_mean_token_accuracy": 0.8005841762678964,
+ "eval_num_tokens": 822969.0,
+ "eval_runtime": 54.1306,
+ "eval_samples_per_second": 25.771,
+ "eval_steps_per_second": 3.233,
+ "step": 320
+ },
+ {
+ "entropy": 0.7627703189849854,
+ "epoch": 0.8348680171884592,
+ "grad_norm": 0.957506537437439,
+ "learning_rate": 6.0940715405257566e-05,
+ "loss": 0.679268217086792,
+ "mean_token_accuracy": 0.8040587857365609,
+ "num_tokens": 873459.0,
+ "step": 340
+ },
+ {
+ "epoch": 0.8348680171884592,
+ "eval_entropy": 0.7624005011149815,
+ "eval_loss": 0.6934819221496582,
+ "eval_mean_token_accuracy": 0.8014607231957572,
+ "eval_num_tokens": 873459.0,
+ "eval_runtime": 54.1656,
+ "eval_samples_per_second": 25.754,
+ "eval_steps_per_second": 3.231,
+ "step": 340
+ },
+ {
+ "entropy": 0.763699171692133,
+ "epoch": 0.8839779005524862,
+ "grad_norm": 1.4329766035079956,
+ "learning_rate": 5.767668664158135e-05,
+ "loss": 0.6764235019683837,
+ "mean_token_accuracy": 0.8051542706787587,
+ "num_tokens": 925669.0,
+ "step": 360
+ },
+ {
+ "epoch": 0.8839779005524862,
+ "eval_entropy": 0.7163350166593279,
+ "eval_loss": 0.6882685422897339,
+ "eval_mean_token_accuracy": 0.802961312021528,
+ "eval_num_tokens": 925669.0,
+ "eval_runtime": 53.8288,
+ "eval_samples_per_second": 25.916,
+ "eval_steps_per_second": 3.251,
+ "step": 360
+ },
+ {
+ "entropy": 0.7710359543561935,
+ "epoch": 0.9330877839165131,
+ "grad_norm": 1.141253113746643,
+ "learning_rate": 5.429729542559189e-05,
+ "loss": 0.6897628784179688,
+ "mean_token_accuracy": 0.8005676351487636,
+ "num_tokens": 973436.0,
+ "step": 380
+ },
+ {
+ "epoch": 0.9330877839165131,
+ "eval_entropy": 0.7462973778588431,
+ "eval_loss": 0.6829195022583008,
+ "eval_mean_token_accuracy": 0.8036447630609784,
+ "eval_num_tokens": 973436.0,
+ "eval_runtime": 53.8276,
+ "eval_samples_per_second": 25.916,
+ "eval_steps_per_second": 3.251,
+ "step": 380
+ },
+ {
+ "entropy": 0.7500453963875771,
+ "epoch": 0.9821976672805403,
+ "grad_norm": 0.9055600166320801,
+ "learning_rate": 5.0827289800936774e-05,
+ "loss": 0.6726306915283203,
+ "mean_token_accuracy": 0.8026990614831447,
+ "num_tokens": 1026370.0,
+ "step": 400
+ },
+ {
+ "epoch": 0.9821976672805403,
+ "eval_entropy": 0.745090776511601,
+ "eval_loss": 0.6795706152915955,
+ "eval_mean_token_accuracy": 0.8038556255613054,
+ "eval_num_tokens": 1026370.0,
+ "eval_runtime": 53.9485,
+ "eval_samples_per_second": 25.858,
+ "eval_steps_per_second": 3.244,
+ "step": 400
+ },
+ {
+ "entropy": 0.7288912821125675,
+ "epoch": 1.0294659300184161,
+ "grad_norm": 0.9075890779495239,
+ "learning_rate": 4.729208140100435e-05,
+ "loss": 0.6533905506134033,
+ "mean_token_accuracy": 0.807745551908171,
+ "num_tokens": 1076273.0,
+ "step": 420
+ },
+ {
+ "epoch": 1.0294659300184161,
+ "eval_entropy": 0.7279435137340001,
+ "eval_loss": 0.6758045554161072,
+ "eval_mean_token_accuracy": 0.8043134975433349,
+ "eval_num_tokens": 1076273.0,
+ "eval_runtime": 53.8173,
+ "eval_samples_per_second": 25.921,
+ "eval_steps_per_second": 3.252,
+ "step": 420
+ },
+ {
+ "entropy": 0.7183651916682721,
+ "epoch": 1.0785758133824432,
+ "grad_norm": 0.9271263480186462,
+ "learning_rate": 4.371755935380415e-05,
+ "loss": 0.6382484912872315,
+ "mean_token_accuracy": 0.810582410544157,
+ "num_tokens": 1133367.0,
+ "step": 440
+ },
+ {
+ "epoch": 1.0785758133824432,
+ "eval_entropy": 0.7224361307280405,
+ "eval_loss": 0.6700165271759033,
+ "eval_mean_token_accuracy": 0.8052176751409258,
+ "eval_num_tokens": 1133367.0,
+ "eval_runtime": 54.2346,
+ "eval_samples_per_second": 25.722,
+ "eval_steps_per_second": 3.227,
+ "step": 440
+ },
+ {
+ "entropy": 0.7097452029585838,
+ "epoch": 1.1276856967464703,
+ "grad_norm": 0.8726329207420349,
+ "learning_rate": 4.012990069004659e-05,
+ "loss": 0.6309147834777832,
+ "mean_token_accuracy": 0.8141021691262722,
+ "num_tokens": 1186585.0,
+ "step": 460
+ },
+ {
+ "epoch": 1.1276856967464703,
+ "eval_entropy": 0.7231041298593793,
+ "eval_loss": 0.6672903299331665,
+ "eval_mean_token_accuracy": 0.8079326885087149,
+ "eval_num_tokens": 1186585.0,
+ "eval_runtime": 53.7923,
+ "eval_samples_per_second": 25.933,
+ "eval_steps_per_second": 3.253,
+ "step": 460
+ },
+ {
+ "entropy": 0.7170382678508759,
+ "epoch": 1.1767955801104972,
+ "grad_norm": 0.8970508575439453,
+ "learning_rate": 3.6555378642846364e-05,
+ "loss": 0.6294277667999267,
+ "mean_token_accuracy": 0.8117733597755432,
+ "num_tokens": 1235457.0,
+ "step": 480
+ },
+ {
+ "epoch": 1.1767955801104972,
+ "eval_entropy": 0.7064940530913216,
+ "eval_loss": 0.6625319123268127,
+ "eval_mean_token_accuracy": 0.8093922696794782,
+ "eval_num_tokens": 1235457.0,
+ "eval_runtime": 53.6574,
+ "eval_samples_per_second": 25.998,
+ "eval_steps_per_second": 3.261,
+ "step": 480
+ },
+ {
+ "entropy": 0.6913388505578041,
+ "epoch": 1.2259054634745243,
+ "grad_norm": 0.9702943563461304,
+ "learning_rate": 3.3020170242913956e-05,
+ "loss": 0.616893196105957,
+ "mean_token_accuracy": 0.8174013815820217,
+ "num_tokens": 1289995.0,
+ "step": 500
+ },
+ {
+ "epoch": 1.2259054634745243,
+ "eval_entropy": 0.7072116715567452,
+ "eval_loss": 0.6600467562675476,
+ "eval_mean_token_accuracy": 0.8099266011374338,
+ "eval_num_tokens": 1289995.0,
+ "eval_runtime": 53.6227,
+ "eval_samples_per_second": 26.015,
+ "eval_steps_per_second": 3.264,
+ "step": 500
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 2.065624617760397e+16,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6d05047585cef4bba4ea582a1f394fe467801e35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 64,
+ "lora_bias": false,
+ "lora_dropout": 0.055708383706168485,
+ "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": 32,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a92660e6fafc54b52681103e85e54bc17d04c26
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/trainer_state.json
@@ -0,0 +1,97 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.14732965009208104,
+ "eval_steps": 20,
+ "global_step": 60,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.5655054792761804,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 2.106731653213501,
+ "learning_rate": 1.9428070010160535e-05,
+ "loss": 2.3814184188842775,
+ "mean_token_accuracy": 0.5490218658000231,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 2.196743746485029,
+ "eval_loss": 2.028722047805786,
+ "eval_mean_token_accuracy": 0.5899124583176204,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 55.8819,
+ "eval_samples_per_second": 24.963,
+ "eval_steps_per_second": 3.132,
+ "step": 20
+ },
+ {
+ "entropy": 1.6433217093348502,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.7636586427688599,
+ "learning_rate": 3.987867002085583e-05,
+ "loss": 1.4883951187133788,
+ "mean_token_accuracy": 0.6520100250840187,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 1.2071456554957798,
+ "eval_loss": 1.1546095609664917,
+ "eval_mean_token_accuracy": 0.7017311205182757,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 55.21,
+ "eval_samples_per_second": 25.267,
+ "eval_steps_per_second": 3.17,
+ "step": 40
+ },
+ {
+ "entropy": 1.1379728086292744,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 1.6891512870788574,
+ "learning_rate": 6.032927003155113e-05,
+ "loss": 1.0574092864990234,
+ "mean_token_accuracy": 0.7190469801425934,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 1.026683736869267,
+ "eval_loss": 0.9847686290740967,
+ "eval_mean_token_accuracy": 0.7383982699257987,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 54.8863,
+ "eval_samples_per_second": 25.416,
+ "eval_steps_per_second": 3.188,
+ "step": 60
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 2542466390807040.0,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/README.md b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+
+
+
+
+## Model Details
+
+### Model Description
+
+
+
+
+
+- **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]
+
+
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+
+
+### Direct Use
+
+
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+
+
+[More Information Needed]
+
+### Recommendations
+
+
+
+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
+
+
+
+[More Information Needed]
+
+### Training Procedure
+
+
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed]
+
+#### Speeds, Sizes, Times [optional]
+
+
+
+[More Information Needed]
+
+## Evaluation
+
+
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+
+
+[More Information Needed]
+
+#### Factors
+
+
+
+[More Information Needed]
+
+#### Metrics
+
+
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+
+
+[More Information Needed]
+
+## Environmental Impact
+
+
+
+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]
+
+
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+
+
+[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
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6d05047585cef4bba4ea582a1f394fe467801e35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/adapter_config.json
@@ -0,0 +1,48 @@
+{
+ "alora_invocation_tokens": null,
+ "alpha_pattern": {},
+ "arrow_config": null,
+ "auto_mapping": null,
+ "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+ "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": 64,
+ "lora_bias": false,
+ "lora_dropout": 0.055708383706168485,
+ "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": 32,
+ "rank_pattern": {},
+ "revision": null,
+ "target_modules": [
+ "k_proj",
+ "gate_proj",
+ "v_proj",
+ "down_proj",
+ "o_proj",
+ "up_proj",
+ "q_proj"
+ ],
+ "target_parameters": null,
+ "task_type": "CAUSAL_LM",
+ "trainable_token_indices": null,
+ "use_bdlora": null,
+ "use_dora": false,
+ "use_qalora": false,
+ "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+ {%- if content is string %}
+ {{- content }}
+ {%- elif content is iterable and content is not mapping %}
+ {%- for item in content %}
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain images.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set image_count.value = image_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+ {%- elif 'video' in item or item.type == 'video' %}
+ {%- if is_system_content %}
+ {{- raise_exception('System message cannot contain videos.') }}
+ {%- endif %}
+ {%- if do_vision_count %}
+ {%- set video_count.value = video_count.value + 1 %}
+ {%- endif %}
+ {%- if add_vision_id %}
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
+ {%- endif %}
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+ {%- elif 'text' in item %}
+ {{- item.text }}
+ {%- else %}
+ {{- raise_exception('Unexpected item type in content.') }}
+ {%- endif %}
+ {%- endfor %}
+ {%- elif content is none or content is undefined %}
+ {{- '' }}
+ {%- else %}
+ {{- raise_exception('Unexpected content type.') }}
+ {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+ {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+ {{- '<|im_start|>system\n' }}
+ {{- "# Tools\n\nYou have access to the following functions:\n\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n" }}
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n\n\n\nvalue_1\n\n\nThis is the value for the second parameter\nthat can span\nmultiple lines\n\n\n\n\n\nReminder:\n- Function calls MUST follow the specified format: an inner block must be nested within XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n' }}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {%- if content %}
+ {{- '\n\n' + content }}
+ {%- endif %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" %}
+ {%- set content = render_content(message.content, false)|trim %}
+ {%- if not(content.startswith('') and content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+ {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+ {%- set content = render_content(message.content, true)|trim %}
+ {%- if message.role == "system" %}
+ {%- if not loop.first %}
+ {{- raise_exception('System message must be at the beginning.') }}
+ {%- endif %}
+ {%- elif message.role == "user" %}
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is string %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in content %}
+ {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- set content = content.split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- set reasoning_content = reasoning_content|trim %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content + '\n\n\n' + content }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if tool_call.function is defined %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {%- if loop.first %}
+ {%- if content|trim %}
+ {{- '\n\n\n\n' }}
+ {%- else %}
+ {{- '\n\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- '\n\n\n' }}
+ {%- endif %}
+ {%- if tool_call.arguments is defined %}
+ {%- for args_name, args_value in tool_call.arguments|items %}
+ {{- '\n' }}
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+ {{- args_value }}
+ {{- '\n\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- content }}
+ {{- '\n' }}
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
+ {{- '<|im_end|>\n' }}
+ {%- elif loop.last %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- else %}
+ {{- raise_exception('Unexpected message role.') }}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is true %}
+ {{- '\n' }}
+ {%- else %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+ "add_prefix_space": false,
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|endoftext|>",
+ "errors": "replace",
+ "image_token": "<|image_pad|>",
+ "is_local": false,
+ "model_max_length": 262144,
+ "model_specific_special_tokens": {
+ "audio_bos_token": "<|audio_start|>",
+ "audio_eos_token": "<|audio_end|>",
+ "audio_token": "<|audio_pad|>",
+ "image_token": "<|image_pad|>",
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+ },
+ "pad_token": "<|endoftext|>",
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+ "split_special_tokens": false,
+ "tokenizer_class": "TokenizersBackend",
+ "unk_token": null,
+ "video_token": "<|video_pad|>",
+ "vision_bos_token": "<|vision_start|>",
+ "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..497471f06721a8c4776e66b7db00f99b473ef627
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-2B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/trainer_state.json
@@ -0,0 +1,118 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 0.19643953345610804,
+ "eval_steps": 20,
+ "global_step": 80,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "entropy": 2.5655054792761804,
+ "epoch": 0.04910988336402701,
+ "grad_norm": 2.106731653213501,
+ "learning_rate": 1.9428070010160535e-05,
+ "loss": 2.3814184188842775,
+ "mean_token_accuracy": 0.5490218658000231,
+ "num_tokens": 51407.0,
+ "step": 20
+ },
+ {
+ "epoch": 0.04910988336402701,
+ "eval_entropy": 2.196743746485029,
+ "eval_loss": 2.028722047805786,
+ "eval_mean_token_accuracy": 0.5899124583176204,
+ "eval_num_tokens": 51407.0,
+ "eval_runtime": 55.8819,
+ "eval_samples_per_second": 24.963,
+ "eval_steps_per_second": 3.132,
+ "step": 20
+ },
+ {
+ "entropy": 1.6433217093348502,
+ "epoch": 0.09821976672805402,
+ "grad_norm": 1.7636586427688599,
+ "learning_rate": 3.987867002085583e-05,
+ "loss": 1.4883951187133788,
+ "mean_token_accuracy": 0.6520100250840187,
+ "num_tokens": 107414.0,
+ "step": 40
+ },
+ {
+ "epoch": 0.09821976672805402,
+ "eval_entropy": 1.2071456554957798,
+ "eval_loss": 1.1546095609664917,
+ "eval_mean_token_accuracy": 0.7017311205182757,
+ "eval_num_tokens": 107414.0,
+ "eval_runtime": 55.21,
+ "eval_samples_per_second": 25.267,
+ "eval_steps_per_second": 3.17,
+ "step": 40
+ },
+ {
+ "entropy": 1.1379728086292744,
+ "epoch": 0.14732965009208104,
+ "grad_norm": 1.6891512870788574,
+ "learning_rate": 6.032927003155113e-05,
+ "loss": 1.0574092864990234,
+ "mean_token_accuracy": 0.7190469801425934,
+ "num_tokens": 157071.0,
+ "step": 60
+ },
+ {
+ "epoch": 0.14732965009208104,
+ "eval_entropy": 1.026683736869267,
+ "eval_loss": 0.9847686290740967,
+ "eval_mean_token_accuracy": 0.7383982699257987,
+ "eval_num_tokens": 157071.0,
+ "eval_runtime": 54.8863,
+ "eval_samples_per_second": 25.416,
+ "eval_steps_per_second": 3.188,
+ "step": 60
+ },
+ {
+ "entropy": 1.0109580919146537,
+ "epoch": 0.19643953345610804,
+ "grad_norm": 1.7396622896194458,
+ "learning_rate": 8.077987004224643e-05,
+ "loss": 0.9192052841186523,
+ "mean_token_accuracy": 0.7471949681639671,
+ "num_tokens": 204521.0,
+ "step": 80
+ },
+ {
+ "epoch": 0.19643953345610804,
+ "eval_entropy": 0.9901534714017596,
+ "eval_loss": 0.8948126435279846,
+ "eval_mean_token_accuracy": 0.7570484624590192,
+ "eval_num_tokens": 204521.0,
+ "eval_runtime": 54.924,
+ "eval_samples_per_second": 25.399,
+ "eval_steps_per_second": 3.186,
+ "step": 80
+ }
+ ],
+ "logging_steps": 20,
+ "max_steps": 816,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 2,
+ "save_steps": 20,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": true,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 3395691794598912.0,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}