Duplicate from vidore/colqwen2-v0.1

Co-authored-by: Manuel Faysse <manu@users.noreply.huggingface.co>

.gitattributes

@@ -0,0 +1,35 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,121 @@
+---
+license: mit
+library_name: colpali
+base_model: vidore/colqwen2-base
+language:
+- en
+tags:
+- colpali
+- vidore
+---
+# ColQwen2: Visual Retriever based on Qwen2-VL-2B-Instruct with ColBERT strategy
+
+ColQwen is a model based on a novel model architecture and training strategy based on Vision Language Models (VLMs) to efficiently index documents from their visual features.
+It is a [Qwen2-VL-2B](https://huggingface.co/Qwen/Qwen2-VL-2B-Instruct) extension that generates [ColBERT](https://arxiv.org/abs/2004.12832)- style multi-vector representations of text and images. 
+It was introduced in the paper [ColPali: Efficient Document Retrieval with Vision Language Models](https://arxiv.org/abs/2407.01449) and first released in [this repository](https://github.com/ManuelFay/colpali)
+
+This version is the untrained base version to guarantee deterministic projection layer initialization.
+<p align="center"><img width=800 src="https://github.com/illuin-tech/colpali/blob/main/assets/colpali_architecture.webp?raw=true"/></p>
+
+## Version specificity
+
+
+This model takes dynamic image resolutions in input and does not resize them, changing their aspect ratio as in ColPali.
+Maximal resolution is set so that 768 image patches are created at most. Experiments show clear improvements with larger amounts of image patches, at the cost of memory requirements.
+
+This version is trained with `colpali-engine==0.3.1`.
+
+Data is the same as the ColPali data described in the paper.
+
+
+## Model Training
+
+### Dataset
+Our training dataset of 127,460 query-page pairs is comprised of train sets of openly available academic datasets (63%) and a synthetic dataset made up of pages from web-crawled PDF documents and augmented with VLM-generated (Claude-3 Sonnet) pseudo-questions (37%). 
+Our training set is fully English by design, enabling us to study zero-shot generalization to non-English languages. We explicitly verify no multi-page PDF document is used both [*ViDoRe*](https://huggingface.co/collections/vidore/vidore-benchmark-667173f98e70a1c0fa4db00d) and in the train set to prevent evaluation contamination. 
+A validation set is created with 2% of the samples to tune hyperparameters.
+
+*Note: Multilingual data is present in the pretraining corpus of the language model and most probably in the multimodal training.*
+
+### Parameters
+
+All models are trained for 1 epoch on the train set. Unless specified otherwise, we train models in `bfloat16` format, use low-rank adapters ([LoRA](https://arxiv.org/abs/2106.09685)) 
+with `alpha=32`  and `r=32` on the transformer layers from the language model, 
+as well as the final randomly initialized projection layer, and use a `paged_adamw_8bit` optimizer. 
+We train on an 8 GPU setup with data parallelism, a learning rate of 5e-5 with linear decay with 2.5% warmup steps, and a batch size of 32.
+
+## Usage
+
+Make sure `colpali-engine` is installed from source or with a version superior to 0.3.1.
+`transformers` version must be > 4.45.0.
+
+```bash
+pip install git+https://github.com/illuin-tech/colpali
+```
+
+```python
+import torch
+from PIL import Image
+
+from colpali_engine.models import ColQwen2, ColQwen2Processor
+
+model = ColQwen2.from_pretrained(
+        "vidore/colqwen2-v0.1",
+        torch_dtype=torch.bfloat16,
+        device_map="cuda:0",  # or "mps" if on Apple Silicon
+    )
+processor = ColQwen2Processor.from_pretrained("vidore/colqwen2-v0.1")
+
+# Your inputs
+images = [
+    Image.new("RGB", (32, 32), color="white"),
+    Image.new("RGB", (16, 16), color="black"),
+]
+queries = [
+    "Is attention really all you need?",
+    "What is the amount of bananas farmed in Salvador?",
+]
+
+# Process the inputs
+batch_images = processor.process_images(images).to(model.device)
+batch_queries = processor.process_queries(queries).to(model.device)
+
+# Forward pass
+with torch.no_grad():
+    image_embeddings = model(**batch_images)
+    query_embeddings = model(**batch_queries)
+
+scores = processor.score_multi_vector(query_embeddings, image_embeddings)
+```
+
+
+## Limitations
+
+ - **Focus**: The model primarily focuses on PDF-type documents and high-ressources languages, potentially limiting its generalization to other document types or less represented languages.
+ - **Support**: The model relies on multi-vector retreiving derived from the ColBERT late interaction mechanism, which may require engineering efforts to adapt to widely used vector retrieval frameworks that lack native multi-vector support.
+
+## License
+
+ColQwen2's vision language backbone model (Qwen2-VL) is under `apache2.0` license. The adapters attached to the model are under MIT license.
+
+## Contact
+
+- Manuel Faysse: manuel.faysse@illuin.tech
+- Hugues Sibille: hugues.sibille@illuin.tech
+- Tony Wu: tony.wu@illuin.tech
+
+## Citation
+
+If you use any datasets or models from this organization in your research, please cite the original dataset as follows:
+
+```bibtex
+@misc{faysse2024colpaliefficientdocumentretrieval,
+  title={ColPali: Efficient Document Retrieval with Vision Language Models}, 
+  author={Manuel Faysse and Hugues Sibille and Tony Wu and Bilel Omrani and Gautier Viaud and Céline Hudelot and Pierre Colombo},
+  year={2024},
+  eprint={2407.01449},
+  archivePrefix={arXiv},
+  primaryClass={cs.IR},
+  url={https://arxiv.org/abs/2407.01449}, 
+}
+```

adapter_config.json

@@ -0,0 +1,26 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "vidore/colqwen2-base",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": "gaussian",
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 32,
+  "lora_dropout": 0.1,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": "(.*(model).*(down_proj|gate_proj|up_proj|k_proj|q_proj|v_proj|o_proj).*$|.*(custom_text_proj).*$)",
+  "task_type": "FEATURE_EXTRACTION",
+  "use_dora": false,
+  "use_rslora": false
+}

adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f26000fd1fbd64be94a32e99012afde7b375f8342bb54b7a7dd0c1dd0d57066
+size 74018232

added_tokens.json

@@ -0,0 +1,16 @@
+{
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

chat_template.json

@@ -0,0 +1,3 @@
+{
+  "chat_template": "{% set image_count = namespace(value=0) %}{% set video_count = namespace(value=0) %}{% for message in messages %}{% if loop.first and message['role'] != 'system' %}<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n{% endif %}<|im_start|>{{ message['role'] }}\n{% if message['content'] is string %}{{ message['content'] }}<|im_end|>\n{% else %}{% for content in message['content'] %}{% if content['type'] == 'image' or 'image' in content or 'image_url' in content %}{% set image_count.value = image_count.value + 1 %}{% if add_vision_id %}Picture {{ image_count.value }}: {% endif %}<|vision_start|><|image_pad|><|vision_end|>{% elif content['type'] == 'video' or 'video' in content %}{% set video_count.value = video_count.value + 1 %}{% if add_vision_id %}Video {{ video_count.value }}: {% endif %}<|vision_start|><|video_pad|><|vision_end|>{% elif 'text' in content %}{{ content['text'] }}{% endif %}{% endfor %}<|im_end|>\n{% endif %}{% endfor %}{% if add_generation_prompt %}<|im_start|>assistant\n{% endif %}"
+}

checkpoint-3694/README.md

@@ -0,0 +1,202 @@
+---
+base_model: ./models/colqwen2_base
+library_name: peft
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.11.1

checkpoint-3694/adapter_config.json

@@ -0,0 +1,26 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "./models/colqwen2_base",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": "gaussian",
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 32,
+  "lora_dropout": 0.1,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": "(.*(model).*(down_proj|gate_proj|up_proj|k_proj|q_proj|v_proj|o_proj).*$|.*(custom_text_proj).*$)",
+  "task_type": "FEATURE_EXTRACTION",
+  "use_dora": false,
+  "use_rslora": false
+}

checkpoint-3694/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f26000fd1fbd64be94a32e99012afde7b375f8342bb54b7a7dd0c1dd0d57066
+size 74018232

checkpoint-3694/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94dfe0075ec4f21387ac18e77c648bfc5db314e088bdcdb8d4fdf6e30532496b
+size 148262384

checkpoint-3694/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:75e9665951c5d547a5de8ed05c5981901bf7eeb728786a0acc4cf999aa58e092
+size 14244

checkpoint-3694/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f0c4c047aa3abb1bc366c2fb0e6a4fd601029da0338c37929023a3ff5f04bef3
+size 1064

checkpoint-3694/trainer_state.json

@@ -0,0 +1,2904 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 100,
+  "global_step": 3694,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0027070925825663237,
+      "grad_norm": 5.65625,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 0.7343,
+      "step": 10
+    },
+    {
+      "epoch": 0.005414185165132647,
+      "grad_norm": 8.9375,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 0.864,
+      "step": 20
+    },
+    {
+      "epoch": 0.008121277747698972,
+      "grad_norm": 9.0625,
+      "learning_rate": 3e-06,
+      "loss": 0.7618,
+      "step": 30
+    },
+    {
+      "epoch": 0.010828370330265295,
+      "grad_norm": 7.375,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.749,
+      "step": 40
+    },
+    {
+      "epoch": 0.01353546291283162,
+      "grad_norm": 5.4375,
+      "learning_rate": 5e-06,
+      "loss": 0.7434,
+      "step": 50
+    },
+    {
+      "epoch": 0.016242555495397944,
+      "grad_norm": 5.34375,
+      "learning_rate": 6e-06,
+      "loss": 0.7079,
+      "step": 60
+    },
+    {
+      "epoch": 0.018949648077964266,
+      "grad_norm": 4.25,
+      "learning_rate": 7.000000000000001e-06,
+      "loss": 0.7368,
+      "step": 70
+    },
+    {
+      "epoch": 0.02165674066053059,
+      "grad_norm": 3.828125,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.682,
+      "step": 80
+    },
+    {
+      "epoch": 0.024363833243096916,
+      "grad_norm": 2.875,
+      "learning_rate": 9e-06,
+      "loss": 0.6489,
+      "step": 90
+    },
+    {
+      "epoch": 0.02707092582566324,
+      "grad_norm": 2.453125,
+      "learning_rate": 1e-05,
+      "loss": 0.589,
+      "step": 100
+    },
+    {
+      "epoch": 0.02707092582566324,
+      "eval_loss": 0.6064066290855408,
+      "eval_runtime": 155.4734,
+      "eval_samples_per_second": 3.293,
+      "eval_steps_per_second": 0.103,
+      "step": 100
+    },
+    {
+      "epoch": 0.02977801840822956,
+      "grad_norm": 2.1875,
+      "learning_rate": 1.1000000000000001e-05,
+      "loss": 0.5723,
+      "step": 110
+    },
+    {
+      "epoch": 0.03248511099079589,
+      "grad_norm": 3.640625,
+      "learning_rate": 1.2e-05,
+      "loss": 0.5525,
+      "step": 120
+    },
+    {
+      "epoch": 0.03519220357336221,
+      "grad_norm": 2.34375,
+      "learning_rate": 1.3000000000000001e-05,
+      "loss": 0.5503,
+      "step": 130
+    },
+    {
+      "epoch": 0.03789929615592853,
+      "grad_norm": 2.3125,
+      "learning_rate": 1.4000000000000001e-05,
+      "loss": 0.4926,
+      "step": 140
+    },
+    {
+      "epoch": 0.040606388738494856,
+      "grad_norm": 2.578125,
+      "learning_rate": 1.5e-05,
+      "loss": 0.4831,
+      "step": 150
+    },
+    {
+      "epoch": 0.04331348132106118,
+      "grad_norm": 2.109375,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.4413,
+      "step": 160
+    },
+    {
+      "epoch": 0.0460205739036275,
+      "grad_norm": 2.671875,
+      "learning_rate": 1.7000000000000003e-05,
+      "loss": 0.4562,
+      "step": 170
+    },
+    {
+      "epoch": 0.04872766648619383,
+      "grad_norm": 2.203125,
+      "learning_rate": 1.8e-05,
+      "loss": 0.4348,
+      "step": 180
+    },
+    {
+      "epoch": 0.051434759068760154,
+      "grad_norm": 2.015625,
+      "learning_rate": 1.9e-05,
+      "loss": 0.4386,
+      "step": 190
+    },
+    {
+      "epoch": 0.05414185165132648,
+      "grad_norm": 2.09375,
+      "learning_rate": 2e-05,
+      "loss": 0.4009,
+      "step": 200
+    },
+    {
+      "epoch": 0.05414185165132648,
+      "eval_loss": 0.4071368873119354,
+      "eval_runtime": 104.5667,
+      "eval_samples_per_second": 4.896,
+      "eval_steps_per_second": 0.153,
+      "step": 200
+    },
+    {
+      "epoch": 0.0568489442338928,
+      "grad_norm": 2.25,
+      "learning_rate": 2.1e-05,
+      "loss": 0.4026,
+      "step": 210
+    },
+    {
+      "epoch": 0.05955603681645912,
+      "grad_norm": 1.90625,
+      "learning_rate": 2.2000000000000003e-05,
+      "loss": 0.3774,
+      "step": 220
+    },
+    {
+      "epoch": 0.062263129399025445,
+      "grad_norm": 1.8828125,
+      "learning_rate": 2.3000000000000003e-05,
+      "loss": 0.3734,
+      "step": 230
+    },
+    {
+      "epoch": 0.06497022198159177,
+      "grad_norm": 1.5703125,
+      "learning_rate": 2.4e-05,
+      "loss": 0.331,
+      "step": 240
+    },
+    {
+      "epoch": 0.0676773145641581,
+      "grad_norm": 1.671875,
+      "learning_rate": 2.5e-05,
+      "loss": 0.3381,
+      "step": 250
+    },
+    {
+      "epoch": 0.07038440714672442,
+      "grad_norm": 1.8515625,
+      "learning_rate": 2.6000000000000002e-05,
+      "loss": 0.3392,
+      "step": 260
+    },
+    {
+      "epoch": 0.07309149972929074,
+      "grad_norm": 1.625,
+      "learning_rate": 2.7000000000000002e-05,
+      "loss": 0.3165,
+      "step": 270
+    },
+    {
+      "epoch": 0.07579859231185707,
+      "grad_norm": 1.671875,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.3162,
+      "step": 280
+    },
+    {
+      "epoch": 0.07850568489442339,
+      "grad_norm": 1.953125,
+      "learning_rate": 2.9e-05,
+      "loss": 0.2844,
+      "step": 290
+    },
+    {
+      "epoch": 0.08121277747698971,
+      "grad_norm": 2.28125,
+      "learning_rate": 3e-05,
+      "loss": 0.3093,
+      "step": 300
+    },
+    {
+      "epoch": 0.08121277747698971,
+      "eval_loss": 0.2839646637439728,
+      "eval_runtime": 103.4698,
+      "eval_samples_per_second": 4.948,
+      "eval_steps_per_second": 0.155,
+      "step": 300
+    },
+    {
+      "epoch": 0.08391987005955603,
+      "grad_norm": 1.4296875,
+      "learning_rate": 3.1e-05,
+      "loss": 0.2753,
+      "step": 310
+    },
+    {
+      "epoch": 0.08662696264212236,
+      "grad_norm": 1.4609375,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.2625,
+      "step": 320
+    },
+    {
+      "epoch": 0.08933405522468868,
+      "grad_norm": 1.5390625,
+      "learning_rate": 3.3e-05,
+      "loss": 0.2525,
+      "step": 330
+    },
+    {
+      "epoch": 0.092041147807255,
+      "grad_norm": 1.9453125,
+      "learning_rate": 3.4000000000000007e-05,
+      "loss": 0.2897,
+      "step": 340
+    },
+    {
+      "epoch": 0.09474824038982133,
+      "grad_norm": 1.4453125,
+      "learning_rate": 3.5e-05,
+      "loss": 0.243,
+      "step": 350
+    },
+    {
+      "epoch": 0.09745533297238766,
+      "grad_norm": 1.6484375,
+      "learning_rate": 3.6e-05,
+      "loss": 0.2442,
+      "step": 360
+    },
+    {
+      "epoch": 0.10016242555495398,
+      "grad_norm": 1.15625,
+      "learning_rate": 3.7e-05,
+      "loss": 0.2139,
+      "step": 370
+    },
+    {
+      "epoch": 0.10286951813752031,
+      "grad_norm": 1.3203125,
+      "learning_rate": 3.8e-05,
+      "loss": 0.2312,
+      "step": 380
+    },
+    {
+      "epoch": 0.10557661072008663,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.9000000000000006e-05,
+      "loss": 0.2202,
+      "step": 390
+    },
+    {
+      "epoch": 0.10828370330265295,
+      "grad_norm": 1.078125,
+      "learning_rate": 4e-05,
+      "loss": 0.2115,
+      "step": 400
+    },
+    {
+      "epoch": 0.10828370330265295,
+      "eval_loss": 0.23121848702430725,
+      "eval_runtime": 134.1955,
+      "eval_samples_per_second": 3.815,
+      "eval_steps_per_second": 0.119,
+      "step": 400
+    },
+    {
+      "epoch": 0.11099079588521928,
+      "grad_norm": 1.828125,
+      "learning_rate": 4.1e-05,
+      "loss": 0.1976,
+      "step": 410
+    },
+    {
+      "epoch": 0.1136978884677856,
+      "grad_norm": 1.0546875,
+      "learning_rate": 4.2e-05,
+      "loss": 0.2182,
+      "step": 420
+    },
+    {
+      "epoch": 0.11640498105035192,
+      "grad_norm": 1.1015625,
+      "learning_rate": 4.3e-05,
+      "loss": 0.1767,
+      "step": 430
+    },
+    {
+      "epoch": 0.11911207363291824,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.2058,
+      "step": 440
+    },
+    {
+      "epoch": 0.12181916621548457,
+      "grad_norm": 2.25,
+      "learning_rate": 4.5e-05,
+      "loss": 0.168,
+      "step": 450
+    },
+    {
+      "epoch": 0.12452625879805089,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.600000000000001e-05,
+      "loss": 0.1728,
+      "step": 460
+    },
+    {
+      "epoch": 0.12723335138061723,
+      "grad_norm": 2.109375,
+      "learning_rate": 4.7e-05,
+      "loss": 0.1927,
+      "step": 470
+    },
+    {
+      "epoch": 0.12994044396318355,
+      "grad_norm": 2.140625,
+      "learning_rate": 4.8e-05,
+      "loss": 0.1433,
+      "step": 480
+    },
+    {
+      "epoch": 0.13264753654574987,
+      "grad_norm": 1.6640625,
+      "learning_rate": 4.9e-05,
+      "loss": 0.1728,
+      "step": 490
+    },
+    {
+      "epoch": 0.1353546291283162,
+      "grad_norm": 2.4375,
+      "learning_rate": 5e-05,
+      "loss": 0.176,
+      "step": 500
+    },
+    {
+      "epoch": 0.1353546291283162,
+      "eval_loss": 0.19780531525611877,
+      "eval_runtime": 102.6349,
+      "eval_samples_per_second": 4.989,
+      "eval_steps_per_second": 0.156,
+      "step": 500
+    },
+    {
+      "epoch": 0.13806172171088252,
+      "grad_norm": 1.8984375,
+      "learning_rate": 4.984345648090169e-05,
+      "loss": 0.1639,
+      "step": 510
+    },
+    {
+      "epoch": 0.14076881429344884,
+      "grad_norm": 1.5078125,
+      "learning_rate": 4.9686912961803384e-05,
+      "loss": 0.1598,
+      "step": 520
+    },
+    {
+      "epoch": 0.14347590687601516,
+      "grad_norm": 5.625,
+      "learning_rate": 4.9530369442705075e-05,
+      "loss": 0.17,
+      "step": 530
+    },
+    {
+      "epoch": 0.1461829994585815,
+      "grad_norm": 2.390625,
+      "learning_rate": 4.9373825923606765e-05,
+      "loss": 0.1601,
+      "step": 540
+    },
+    {
+      "epoch": 0.1488900920411478,
+      "grad_norm": 2.28125,
+      "learning_rate": 4.9217282404508456e-05,
+      "loss": 0.2067,
+      "step": 550
+    },
+    {
+      "epoch": 0.15159718462371413,
+      "grad_norm": 1.296875,
+      "learning_rate": 4.906073888541015e-05,
+      "loss": 0.1396,
+      "step": 560
+    },
+    {
+      "epoch": 0.15430427720628045,
+      "grad_norm": 2.3125,
+      "learning_rate": 4.890419536631184e-05,
+      "loss": 0.1379,
+      "step": 570
+    },
+    {
+      "epoch": 0.15701136978884678,
+      "grad_norm": 1.984375,
+      "learning_rate": 4.874765184721353e-05,
+      "loss": 0.1722,
+      "step": 580
+    },
+    {
+      "epoch": 0.1597184623714131,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.859110832811522e-05,
+      "loss": 0.1608,
+      "step": 590
+    },
+    {
+      "epoch": 0.16242555495397942,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.843456480901691e-05,
+      "loss": 0.1576,
+      "step": 600
+    },
+    {
+      "epoch": 0.16242555495397942,
+      "eval_loss": 0.1739654242992401,
+      "eval_runtime": 102.4755,
+      "eval_samples_per_second": 4.996,
+      "eval_steps_per_second": 0.156,
+      "step": 600
+    },
+    {
+      "epoch": 0.16513264753654575,
+      "grad_norm": 1.78125,
+      "learning_rate": 4.82780212899186e-05,
+      "loss": 0.1292,
+      "step": 610
+    },
+    {
+      "epoch": 0.16783974011911207,
+      "grad_norm": 2.296875,
+      "learning_rate": 4.812147777082029e-05,
+      "loss": 0.1421,
+      "step": 620
+    },
+    {
+      "epoch": 0.1705468327016784,
+      "grad_norm": 3.53125,
+      "learning_rate": 4.796493425172198e-05,
+      "loss": 0.1473,
+      "step": 630
+    },
+    {
+      "epoch": 0.17325392528424471,
+      "grad_norm": 1.6015625,
+      "learning_rate": 4.780839073262367e-05,
+      "loss": 0.1313,
+      "step": 640
+    },
+    {
+      "epoch": 0.17596101786681104,
+      "grad_norm": 2.75,
+      "learning_rate": 4.765184721352536e-05,
+      "loss": 0.0987,
+      "step": 650
+    },
+    {
+      "epoch": 0.17866811044937736,
+      "grad_norm": 2.453125,
+      "learning_rate": 4.7495303694427054e-05,
+      "loss": 0.1301,
+      "step": 660
+    },
+    {
+      "epoch": 0.18137520303194368,
+      "grad_norm": 1.203125,
+      "learning_rate": 4.7338760175328744e-05,
+      "loss": 0.118,
+      "step": 670
+    },
+    {
+      "epoch": 0.18408229561451,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.7182216656230435e-05,
+      "loss": 0.1384,
+      "step": 680
+    },
+    {
+      "epoch": 0.18678938819707633,
+      "grad_norm": 0.78515625,
+      "learning_rate": 4.7025673137132126e-05,
+      "loss": 0.0855,
+      "step": 690
+    },
+    {
+      "epoch": 0.18949648077964265,
+      "grad_norm": 2.34375,
+      "learning_rate": 4.6869129618033816e-05,
+      "loss": 0.148,
+      "step": 700
+    },
+    {
+      "epoch": 0.18949648077964265,
+      "eval_loss": 0.1687919795513153,
+      "eval_runtime": 102.502,
+      "eval_samples_per_second": 4.995,
+      "eval_steps_per_second": 0.156,
+      "step": 700
+    },
+    {
+      "epoch": 0.19220357336220897,
+      "grad_norm": 0.81640625,
+      "learning_rate": 4.671258609893551e-05,
+      "loss": 0.134,
+      "step": 710
+    },
+    {
+      "epoch": 0.19491066594477532,
+      "grad_norm": 1.3046875,
+      "learning_rate": 4.65560425798372e-05,
+      "loss": 0.1222,
+      "step": 720
+    },
+    {
+      "epoch": 0.19761775852734165,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.639949906073889e-05,
+      "loss": 0.1511,
+      "step": 730
+    },
+    {
+      "epoch": 0.20032485110990797,
+      "grad_norm": 0.55078125,
+      "learning_rate": 4.624295554164057e-05,
+      "loss": 0.1661,
+      "step": 740
+    },
+    {
+      "epoch": 0.2030319436924743,
+      "grad_norm": 1.9140625,
+      "learning_rate": 4.608641202254227e-05,
+      "loss": 0.1381,
+      "step": 750
+    },
+    {
+      "epoch": 0.20573903627504062,
+      "grad_norm": 2.015625,
+      "learning_rate": 4.5929868503443954e-05,
+      "loss": 0.1499,
+      "step": 760
+    },
+    {
+      "epoch": 0.20844612885760694,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.577332498434565e-05,
+      "loss": 0.1692,
+      "step": 770
+    },
+    {
+      "epoch": 0.21115322144017326,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.561678146524734e-05,
+      "loss": 0.1226,
+      "step": 780
+    },
+    {
+      "epoch": 0.21386031402273958,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.546023794614903e-05,
+      "loss": 0.1594,
+      "step": 790
+    },
+    {
+      "epoch": 0.2165674066053059,
+      "grad_norm": 0.76171875,
+      "learning_rate": 4.5303694427050724e-05,
+      "loss": 0.0962,
+      "step": 800
+    },
+    {
+      "epoch": 0.2165674066053059,
+      "eval_loss": 0.15033257007598877,
+      "eval_runtime": 102.8567,
+      "eval_samples_per_second": 4.978,
+      "eval_steps_per_second": 0.156,
+      "step": 800
+    },
+    {
+      "epoch": 0.21927449918787223,
+      "grad_norm": 1.4453125,
+      "learning_rate": 4.5147150907952414e-05,
+      "loss": 0.1581,
+      "step": 810
+    },
+    {
+      "epoch": 0.22198159177043855,
+      "grad_norm": 2.078125,
+      "learning_rate": 4.4990607388854105e-05,
+      "loss": 0.155,
+      "step": 820
+    },
+    {
+      "epoch": 0.22468868435300487,
+      "grad_norm": 0.921875,
+      "learning_rate": 4.4834063869755796e-05,
+      "loss": 0.1103,
+      "step": 830
+    },
+    {
+      "epoch": 0.2273957769355712,
+      "grad_norm": 0.90234375,
+      "learning_rate": 4.4677520350657486e-05,
+      "loss": 0.1323,
+      "step": 840
+    },
+    {
+      "epoch": 0.23010286951813752,
+      "grad_norm": 0.953125,
+      "learning_rate": 4.452097683155918e-05,
+      "loss": 0.1073,
+      "step": 850
+    },
+    {
+      "epoch": 0.23280996210070384,
+      "grad_norm": 0.69921875,
+      "learning_rate": 4.436443331246087e-05,
+      "loss": 0.1333,
+      "step": 860
+    },
+    {
+      "epoch": 0.23551705468327017,
+      "grad_norm": 4.34375,
+      "learning_rate": 4.420788979336256e-05,
+      "loss": 0.1429,
+      "step": 870
+    },
+    {
+      "epoch": 0.2382241472658365,
+      "grad_norm": 0.5859375,
+      "learning_rate": 4.405134627426425e-05,
+      "loss": 0.0988,
+      "step": 880
+    },
+    {
+      "epoch": 0.2409312398484028,
+      "grad_norm": 1.3046875,
+      "learning_rate": 4.389480275516594e-05,
+      "loss": 0.1184,
+      "step": 890
+    },
+    {
+      "epoch": 0.24363833243096913,
+      "grad_norm": 1.0234375,
+      "learning_rate": 4.373825923606763e-05,
+      "loss": 0.118,
+      "step": 900
+    },
+    {
+      "epoch": 0.24363833243096913,
+      "eval_loss": 0.14425741136074066,
+      "eval_runtime": 115.407,
+      "eval_samples_per_second": 4.436,
+      "eval_steps_per_second": 0.139,
+      "step": 900
+    },
+    {
+      "epoch": 0.24634542501353546,
+      "grad_norm": 1.34375,
+      "learning_rate": 4.358171571696932e-05,
+      "loss": 0.1032,
+      "step": 910
+    },
+    {
+      "epoch": 0.24905251759610178,
+      "grad_norm": 1.78125,
+      "learning_rate": 4.342517219787101e-05,
+      "loss": 0.1242,
+      "step": 920
+    },
+    {
+      "epoch": 0.2517596101786681,
+      "grad_norm": 1.296875,
+      "learning_rate": 4.3268628678772696e-05,
+      "loss": 0.131,
+      "step": 930
+    },
+    {
+      "epoch": 0.25446670276123445,
+      "grad_norm": 1.90625,
+      "learning_rate": 4.3112085159674393e-05,
+      "loss": 0.1058,
+      "step": 940
+    },
+    {
+      "epoch": 0.25717379534380075,
+      "grad_norm": 1.1015625,
+      "learning_rate": 4.295554164057608e-05,
+      "loss": 0.155,
+      "step": 950
+    },
+    {
+      "epoch": 0.2598808879263671,
+      "grad_norm": 1.2734375,
+      "learning_rate": 4.2798998121477775e-05,
+      "loss": 0.1103,
+      "step": 960
+    },
+    {
+      "epoch": 0.2625879805089334,
+      "grad_norm": 4.59375,
+      "learning_rate": 4.264245460237946e-05,
+      "loss": 0.1288,
+      "step": 970
+    },
+    {
+      "epoch": 0.26529507309149974,
+      "grad_norm": 0.90234375,
+      "learning_rate": 4.2485911083281156e-05,
+      "loss": 0.0967,
+      "step": 980
+    },
+    {
+      "epoch": 0.26800216567406604,
+      "grad_norm": 1.734375,
+      "learning_rate": 4.232936756418284e-05,
+      "loss": 0.1523,
+      "step": 990
+    },
+    {
+      "epoch": 0.2707092582566324,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.217282404508454e-05,
+      "loss": 0.1364,
+      "step": 1000
+    },
+    {
+      "epoch": 0.2707092582566324,
+      "eval_loss": 0.14593610167503357,
+      "eval_runtime": 104.6862,
+      "eval_samples_per_second": 4.891,
+      "eval_steps_per_second": 0.153,
+      "step": 1000
+    },
+    {
+      "epoch": 0.2734163508391987,
+      "grad_norm": 2.125,
+      "learning_rate": 4.201628052598622e-05,
+      "loss": 0.1228,
+      "step": 1010
+    },
+    {
+      "epoch": 0.27612344342176504,
+      "grad_norm": 1.7265625,
+      "learning_rate": 4.185973700688792e-05,
+      "loss": 0.1114,
+      "step": 1020
+    },
+    {
+      "epoch": 0.27883053600433133,
+      "grad_norm": 1.0859375,
+      "learning_rate": 4.170319348778961e-05,
+      "loss": 0.1016,
+      "step": 1030
+    },
+    {
+      "epoch": 0.2815376285868977,
+      "grad_norm": 0.69921875,
+      "learning_rate": 4.15466499686913e-05,
+      "loss": 0.1023,
+      "step": 1040
+    },
+    {
+      "epoch": 0.284244721169464,
+      "grad_norm": 1.5859375,
+      "learning_rate": 4.139010644959299e-05,
+      "loss": 0.1097,
+      "step": 1050
+    },
+    {
+      "epoch": 0.2869518137520303,
+      "grad_norm": 0.5859375,
+      "learning_rate": 4.123356293049468e-05,
+      "loss": 0.1522,
+      "step": 1060
+    },
+    {
+      "epoch": 0.2896589063345966,
+      "grad_norm": 1.296875,
+      "learning_rate": 4.107701941139637e-05,
+      "loss": 0.1218,
+      "step": 1070
+    },
+    {
+      "epoch": 0.292365998917163,
+      "grad_norm": 1.9765625,
+      "learning_rate": 4.092047589229806e-05,
+      "loss": 0.1491,
+      "step": 1080
+    },
+    {
+      "epoch": 0.29507309149972927,
+      "grad_norm": 0.82421875,
+      "learning_rate": 4.0763932373199754e-05,
+      "loss": 0.1103,
+      "step": 1090
+    },
+    {
+      "epoch": 0.2977801840822956,
+      "grad_norm": 1.96875,
+      "learning_rate": 4.0607388854101445e-05,
+      "loss": 0.1419,
+      "step": 1100
+    },
+    {
+      "epoch": 0.2977801840822956,
+      "eval_loss": 0.14324142038822174,
+      "eval_runtime": 105.9997,
+      "eval_samples_per_second": 4.83,
+      "eval_steps_per_second": 0.151,
+      "step": 1100
+    },
+    {
+      "epoch": 0.3004872766648619,
+      "grad_norm": 0.8984375,
+      "learning_rate": 4.0450845335003135e-05,
+      "loss": 0.0784,
+      "step": 1110
+    },
+    {
+      "epoch": 0.30319436924742826,
+      "grad_norm": 0.9296875,
+      "learning_rate": 4.029430181590482e-05,
+      "loss": 0.0869,
+      "step": 1120
+    },
+    {
+      "epoch": 0.30590146182999456,
+      "grad_norm": 0.6015625,
+      "learning_rate": 4.013775829680652e-05,
+      "loss": 0.1135,
+      "step": 1130
+    },
+    {
+      "epoch": 0.3086085544125609,
+      "grad_norm": 1.4453125,
+      "learning_rate": 3.99812147777082e-05,
+      "loss": 0.1296,
+      "step": 1140
+    },
+    {
+      "epoch": 0.31131564699512726,
+      "grad_norm": 0.8203125,
+      "learning_rate": 3.98246712586099e-05,
+      "loss": 0.122,
+      "step": 1150
+    },
+    {
+      "epoch": 0.31402273957769355,
+      "grad_norm": 1.3203125,
+      "learning_rate": 3.966812773951158e-05,
+      "loss": 0.1102,
+      "step": 1160
+    },
+    {
+      "epoch": 0.3167298321602599,
+      "grad_norm": 0.98046875,
+      "learning_rate": 3.951158422041328e-05,
+      "loss": 0.1284,
+      "step": 1170
+    },
+    {
+      "epoch": 0.3194369247428262,
+      "grad_norm": 2.390625,
+      "learning_rate": 3.9355040701314964e-05,
+      "loss": 0.139,
+      "step": 1180
+    },
+    {
+      "epoch": 0.32214401732539255,
+      "grad_norm": 1.03125,
+      "learning_rate": 3.919849718221666e-05,
+      "loss": 0.1405,
+      "step": 1190
+    },
+    {
+      "epoch": 0.32485110990795885,
+      "grad_norm": 0.75390625,
+      "learning_rate": 3.9041953663118345e-05,
+      "loss": 0.1359,
+      "step": 1200
+    },
+    {
+      "epoch": 0.32485110990795885,
+      "eval_loss": 0.13801878690719604,
+      "eval_runtime": 105.4772,
+      "eval_samples_per_second": 4.854,
+      "eval_steps_per_second": 0.152,
+      "step": 1200
+    },
+    {
+      "epoch": 0.3275582024905252,
+      "grad_norm": 0.9609375,
+      "learning_rate": 3.888541014402004e-05,
+      "loss": 0.0934,
+      "step": 1210
+    },
+    {
+      "epoch": 0.3302652950730915,
+      "grad_norm": 1.9921875,
+      "learning_rate": 3.8728866624921726e-05,
+      "loss": 0.1272,
+      "step": 1220
+    },
+    {
+      "epoch": 0.33297238765565784,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.8572323105823424e-05,
+      "loss": 0.1099,
+      "step": 1230
+    },
+    {
+      "epoch": 0.33567948023822414,
+      "grad_norm": 1.1171875,
+      "learning_rate": 3.841577958672511e-05,
+      "loss": 0.1019,
+      "step": 1240
+    },
+    {
+      "epoch": 0.3383865728207905,
+      "grad_norm": 1.1171875,
+      "learning_rate": 3.8259236067626805e-05,
+      "loss": 0.1503,
+      "step": 1250
+    },
+    {
+      "epoch": 0.3410936654033568,
+      "grad_norm": 0.490234375,
+      "learning_rate": 3.810269254852849e-05,
+      "loss": 0.0857,
+      "step": 1260
+    },
+    {
+      "epoch": 0.34380075798592313,
+      "grad_norm": 0.69140625,
+      "learning_rate": 3.794614902943019e-05,
+      "loss": 0.0949,
+      "step": 1270
+    },
+    {
+      "epoch": 0.34650785056848943,
+      "grad_norm": 2.015625,
+      "learning_rate": 3.778960551033187e-05,
+      "loss": 0.1242,
+      "step": 1280
+    },
+    {
+      "epoch": 0.3492149431510558,
+      "grad_norm": 0.5546875,
+      "learning_rate": 3.763306199123356e-05,
+      "loss": 0.1041,
+      "step": 1290
+    },
+    {
+      "epoch": 0.3519220357336221,
+      "grad_norm": 1.5234375,
+      "learning_rate": 3.747651847213526e-05,
+      "loss": 0.0727,
+      "step": 1300
+    },
+    {
+      "epoch": 0.3519220357336221,
+      "eval_loss": 0.13924407958984375,
+      "eval_runtime": 103.9612,
+      "eval_samples_per_second": 4.925,
+      "eval_steps_per_second": 0.154,
+      "step": 1300
+    },
+    {
+      "epoch": 0.3546291283161884,
+      "grad_norm": 1.0625,
+      "learning_rate": 3.731997495303694e-05,
+      "loss": 0.0898,
+      "step": 1310
+    },
+    {
+      "epoch": 0.3573362208987547,
+      "grad_norm": 2.609375,
+      "learning_rate": 3.716343143393864e-05,
+      "loss": 0.1697,
+      "step": 1320
+    },
+    {
+      "epoch": 0.36004331348132107,
+      "grad_norm": 1.1015625,
+      "learning_rate": 3.7006887914840324e-05,
+      "loss": 0.1111,
+      "step": 1330
+    },
+    {
+      "epoch": 0.36275040606388737,
+      "grad_norm": 0.44921875,
+      "learning_rate": 3.685034439574202e-05,
+      "loss": 0.1194,
+      "step": 1340
+    },
+    {
+      "epoch": 0.3654574986464537,
+      "grad_norm": 1.421875,
+      "learning_rate": 3.6693800876643706e-05,
+      "loss": 0.1019,
+      "step": 1350
+    },
+    {
+      "epoch": 0.36816459122902,
+      "grad_norm": 1.25,
+      "learning_rate": 3.65372573575454e-05,
+      "loss": 0.0984,
+      "step": 1360
+    },
+    {
+      "epoch": 0.37087168381158636,
+      "grad_norm": 0.86328125,
+      "learning_rate": 3.638071383844709e-05,
+      "loss": 0.1298,
+      "step": 1370
+    },
+    {
+      "epoch": 0.37357877639415266,
+      "grad_norm": 1.6484375,
+      "learning_rate": 3.6224170319348784e-05,
+      "loss": 0.1242,
+      "step": 1380
+    },
+    {
+      "epoch": 0.376285868976719,
+      "grad_norm": 0.72265625,
+      "learning_rate": 3.606762680025047e-05,
+      "loss": 0.094,
+      "step": 1390
+    },
+    {
+      "epoch": 0.3789929615592853,
+      "grad_norm": 1.2421875,
+      "learning_rate": 3.5911083281152166e-05,
+      "loss": 0.1286,
+      "step": 1400
+    },
+    {
+      "epoch": 0.3789929615592853,
+      "eval_loss": 0.13358080387115479,
+      "eval_runtime": 120.0689,
+      "eval_samples_per_second": 4.264,
+      "eval_steps_per_second": 0.133,
+      "step": 1400
+    },
+    {
+      "epoch": 0.38170005414185165,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.575453976205385e-05,
+      "loss": 0.1594,
+      "step": 1410
+    },
+    {
+      "epoch": 0.38440714672441795,
+      "grad_norm": 2.0,
+      "learning_rate": 3.559799624295555e-05,
+      "loss": 0.0771,
+      "step": 1420
+    },
+    {
+      "epoch": 0.3871142393069843,
+      "grad_norm": 2.328125,
+      "learning_rate": 3.544145272385723e-05,
+      "loss": 0.1485,
+      "step": 1430
+    },
+    {
+      "epoch": 0.38982133188955065,
+      "grad_norm": 0.30859375,
+      "learning_rate": 3.528490920475893e-05,
+      "loss": 0.0938,
+      "step": 1440
+    },
+    {
+      "epoch": 0.39252842447211694,
+      "grad_norm": 0.91015625,
+      "learning_rate": 3.512836568566061e-05,
+      "loss": 0.1372,
+      "step": 1450
+    },
+    {
+      "epoch": 0.3952355170546833,
+      "grad_norm": 0.6953125,
+      "learning_rate": 3.497182216656231e-05,
+      "loss": 0.1033,
+      "step": 1460
+    },
+    {
+      "epoch": 0.3979426096372496,
+      "grad_norm": 1.0546875,
+      "learning_rate": 3.4815278647463994e-05,
+      "loss": 0.1345,
+      "step": 1470
+    },
+    {
+      "epoch": 0.40064970221981594,
+      "grad_norm": 0.91015625,
+      "learning_rate": 3.4658735128365685e-05,
+      "loss": 0.103,
+      "step": 1480
+    },
+    {
+      "epoch": 0.40335679480238223,
+      "grad_norm": 1.125,
+      "learning_rate": 3.4502191609267375e-05,
+      "loss": 0.083,
+      "step": 1490
+    },
+    {
+      "epoch": 0.4060638873849486,
+      "grad_norm": 0.9296875,
+      "learning_rate": 3.4345648090169066e-05,
+      "loss": 0.1024,
+      "step": 1500
+    },
+    {
+      "epoch": 0.4060638873849486,
+      "eval_loss": 0.13767841458320618,
+      "eval_runtime": 104.1987,
+      "eval_samples_per_second": 4.914,
+      "eval_steps_per_second": 0.154,
+      "step": 1500
+    },
+    {
+      "epoch": 0.4087709799675149,
+      "grad_norm": 1.71875,
+      "learning_rate": 3.418910457107076e-05,
+      "loss": 0.1319,
+      "step": 1510
+    },
+    {
+      "epoch": 0.41147807255008123,
+      "grad_norm": 1.0703125,
+      "learning_rate": 3.403256105197245e-05,
+      "loss": 0.1152,
+      "step": 1520
+    },
+    {
+      "epoch": 0.4141851651326475,
+      "grad_norm": 1.3125,
+      "learning_rate": 3.387601753287414e-05,
+      "loss": 0.1072,
+      "step": 1530
+    },
+    {
+      "epoch": 0.4168922577152139,
+      "grad_norm": 2.171875,
+      "learning_rate": 3.371947401377583e-05,
+      "loss": 0.1408,
+      "step": 1540
+    },
+    {
+      "epoch": 0.41959935029778017,
+      "grad_norm": 1.53125,
+      "learning_rate": 3.3562930494677526e-05,
+      "loss": 0.1125,
+      "step": 1550
+    },
+    {
+      "epoch": 0.4223064428803465,
+      "grad_norm": 1.578125,
+      "learning_rate": 3.340638697557921e-05,
+      "loss": 0.1144,
+      "step": 1560
+    },
+    {
+      "epoch": 0.4250135354629128,
+      "grad_norm": 1.6015625,
+      "learning_rate": 3.324984345648091e-05,
+      "loss": 0.1092,
+      "step": 1570
+    },
+    {
+      "epoch": 0.42772062804547917,
+      "grad_norm": 0.408203125,
+      "learning_rate": 3.309329993738259e-05,
+      "loss": 0.1271,
+      "step": 1580
+    },
+    {
+      "epoch": 0.43042772062804546,
+      "grad_norm": 0.72265625,
+      "learning_rate": 3.293675641828429e-05,
+      "loss": 0.0882,
+      "step": 1590
+    },
+    {
+      "epoch": 0.4331348132106118,
+      "grad_norm": 0.296875,
+      "learning_rate": 3.278021289918597e-05,
+      "loss": 0.1141,
+      "step": 1600
+    },
+    {
+      "epoch": 0.4331348132106118,
+      "eval_loss": 0.13831546902656555,
+      "eval_runtime": 104.916,
+      "eval_samples_per_second": 4.88,
+      "eval_steps_per_second": 0.153,
+      "step": 1600
+    },
+    {
+      "epoch": 0.4358419057931781,
+      "grad_norm": 0.462890625,
+      "learning_rate": 3.262366938008767e-05,
+      "loss": 0.0874,
+      "step": 1610
+    },
+    {
+      "epoch": 0.43854899837574446,
+      "grad_norm": 1.921875,
+      "learning_rate": 3.2467125860989355e-05,
+      "loss": 0.1288,
+      "step": 1620
+    },
+    {
+      "epoch": 0.44125609095831075,
+      "grad_norm": 1.3359375,
+      "learning_rate": 3.231058234189105e-05,
+      "loss": 0.1171,
+      "step": 1630
+    },
+    {
+      "epoch": 0.4439631835408771,
+      "grad_norm": 0.447265625,
+      "learning_rate": 3.2154038822792736e-05,
+      "loss": 0.0845,
+      "step": 1640
+    },
+    {
+      "epoch": 0.4466702761234434,
+      "grad_norm": 1.703125,
+      "learning_rate": 3.1997495303694433e-05,
+      "loss": 0.1052,
+      "step": 1650
+    },
+    {
+      "epoch": 0.44937736870600975,
+      "grad_norm": 1.484375,
+      "learning_rate": 3.184095178459612e-05,
+      "loss": 0.1187,
+      "step": 1660
+    },
+    {
+      "epoch": 0.45208446128857604,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.168440826549781e-05,
+      "loss": 0.1051,
+      "step": 1670
+    },
+    {
+      "epoch": 0.4547915538711424,
+      "grad_norm": 1.9765625,
+      "learning_rate": 3.15278647463995e-05,
+      "loss": 0.1415,
+      "step": 1680
+    },
+    {
+      "epoch": 0.4574986464537087,
+      "grad_norm": 1.3828125,
+      "learning_rate": 3.137132122730119e-05,
+      "loss": 0.1313,
+      "step": 1690
+    },
+    {
+      "epoch": 0.46020573903627504,
+      "grad_norm": 0.8359375,
+      "learning_rate": 3.121477770820288e-05,
+      "loss": 0.1129,
+      "step": 1700
+    },
+    {
+      "epoch": 0.46020573903627504,
+      "eval_loss": 0.13771241903305054,
+      "eval_runtime": 104.8503,
+      "eval_samples_per_second": 4.883,
+      "eval_steps_per_second": 0.153,
+      "step": 1700
+    },
+    {
+      "epoch": 0.4629128316188414,
+      "grad_norm": 1.8359375,
+      "learning_rate": 3.105823418910457e-05,
+      "loss": 0.1618,
+      "step": 1710
+    },
+    {
+      "epoch": 0.4656199242014077,
+      "grad_norm": 0.53515625,
+      "learning_rate": 3.090169067000626e-05,
+      "loss": 0.1025,
+      "step": 1720
+    },
+    {
+      "epoch": 0.46832701678397404,
+      "grad_norm": 4.0625,
+      "learning_rate": 3.074514715090795e-05,
+      "loss": 0.1253,
+      "step": 1730
+    },
+    {
+      "epoch": 0.47103410936654033,
+      "grad_norm": 2.25,
+      "learning_rate": 3.058860363180964e-05,
+      "loss": 0.1098,
+      "step": 1740
+    },
+    {
+      "epoch": 0.4737412019491067,
+      "grad_norm": 1.484375,
+      "learning_rate": 3.0432060112711337e-05,
+      "loss": 0.0868,
+      "step": 1750
+    },
+    {
+      "epoch": 0.476448294531673,
+      "grad_norm": 1.9375,
+      "learning_rate": 3.0275516593613024e-05,
+      "loss": 0.1427,
+      "step": 1760
+    },
+    {
+      "epoch": 0.47915538711423933,
+      "grad_norm": 2.546875,
+      "learning_rate": 3.011897307451472e-05,
+      "loss": 0.1272,
+      "step": 1770
+    },
+    {
+      "epoch": 0.4818624796968056,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.9962429555416406e-05,
+      "loss": 0.16,
+      "step": 1780
+    },
+    {
+      "epoch": 0.484569572279372,
+      "grad_norm": 1.2578125,
+      "learning_rate": 2.9805886036318097e-05,
+      "loss": 0.1079,
+      "step": 1790
+    },
+    {
+      "epoch": 0.48727666486193827,
+      "grad_norm": 0.75,
+      "learning_rate": 2.9649342517219787e-05,
+      "loss": 0.0907,
+      "step": 1800
+    },
+    {
+      "epoch": 0.48727666486193827,
+      "eval_loss": 0.14291706681251526,
+      "eval_runtime": 105.4966,
+      "eval_samples_per_second": 4.853,
+      "eval_steps_per_second": 0.152,
+      "step": 1800
+    },
+    {
+      "epoch": 0.4899837574445046,
+      "grad_norm": 0.94140625,
+      "learning_rate": 2.9492798998121478e-05,
+      "loss": 0.1137,
+      "step": 1810
+    },
+    {
+      "epoch": 0.4926908500270709,
+      "grad_norm": 1.796875,
+      "learning_rate": 2.9336255479023172e-05,
+      "loss": 0.114,
+      "step": 1820
+    },
+    {
+      "epoch": 0.49539794260963727,
+      "grad_norm": 1.4296875,
+      "learning_rate": 2.917971195992486e-05,
+      "loss": 0.1263,
+      "step": 1830
+    },
+    {
+      "epoch": 0.49810503519220356,
+      "grad_norm": 2.171875,
+      "learning_rate": 2.9023168440826553e-05,
+      "loss": 0.1095,
+      "step": 1840
+    },
+    {
+      "epoch": 0.5008121277747699,
+      "grad_norm": 1.2734375,
+      "learning_rate": 2.886662492172824e-05,
+      "loss": 0.1609,
+      "step": 1850
+    },
+    {
+      "epoch": 0.5035192203573362,
+      "grad_norm": 1.84375,
+      "learning_rate": 2.8710081402629935e-05,
+      "loss": 0.1079,
+      "step": 1860
+    },
+    {
+      "epoch": 0.5062263129399025,
+      "grad_norm": 0.328125,
+      "learning_rate": 2.8553537883531622e-05,
+      "loss": 0.1114,
+      "step": 1870
+    },
+    {
+      "epoch": 0.5089334055224689,
+      "grad_norm": 1.9140625,
+      "learning_rate": 2.8396994364433316e-05,
+      "loss": 0.1163,
+      "step": 1880
+    },
+    {
+      "epoch": 0.5116404981050352,
+      "grad_norm": 0.68359375,
+      "learning_rate": 2.8240450845335004e-05,
+      "loss": 0.105,
+      "step": 1890
+    },
+    {
+      "epoch": 0.5143475906876015,
+      "grad_norm": 0.76171875,
+      "learning_rate": 2.8083907326236698e-05,
+      "loss": 0.0931,
+      "step": 1900
+    },
+    {
+      "epoch": 0.5143475906876015,
+      "eval_loss": 0.1378411054611206,
+      "eval_runtime": 122.4527,
+      "eval_samples_per_second": 4.181,
+      "eval_steps_per_second": 0.131,
+      "step": 1900
+    },
+    {
+      "epoch": 0.5170546832701678,
+      "grad_norm": 2.125,
+      "learning_rate": 2.7927363807138385e-05,
+      "loss": 0.1137,
+      "step": 1910
+    },
+    {
+      "epoch": 0.5197617758527342,
+      "grad_norm": 1.265625,
+      "learning_rate": 2.777082028804008e-05,
+      "loss": 0.1127,
+      "step": 1920
+    },
+    {
+      "epoch": 0.5224688684353005,
+      "grad_norm": 1.7734375,
+      "learning_rate": 2.7614276768941766e-05,
+      "loss": 0.1082,
+      "step": 1930
+    },
+    {
+      "epoch": 0.5251759610178668,
+      "grad_norm": 1.6640625,
+      "learning_rate": 2.745773324984346e-05,
+      "loss": 0.1147,
+      "step": 1940
+    },
+    {
+      "epoch": 0.5278830536004331,
+      "grad_norm": 1.84375,
+      "learning_rate": 2.7301189730745148e-05,
+      "loss": 0.1027,
+      "step": 1950
+    },
+    {
+      "epoch": 0.5305901461829995,
+      "grad_norm": 0.86328125,
+      "learning_rate": 2.7144646211646842e-05,
+      "loss": 0.108,
+      "step": 1960
+    },
+    {
+      "epoch": 0.5332972387655658,
+      "grad_norm": 0.59375,
+      "learning_rate": 2.698810269254853e-05,
+      "loss": 0.0917,
+      "step": 1970
+    },
+    {
+      "epoch": 0.5360043313481321,
+      "grad_norm": 0.515625,
+      "learning_rate": 2.683155917345022e-05,
+      "loss": 0.078,
+      "step": 1980
+    },
+    {
+      "epoch": 0.5387114239306985,
+      "grad_norm": 1.1796875,
+      "learning_rate": 2.667501565435191e-05,
+      "loss": 0.0845,
+      "step": 1990
+    },
+    {
+      "epoch": 0.5414185165132648,
+      "grad_norm": 1.7890625,
+      "learning_rate": 2.65184721352536e-05,
+      "loss": 0.0888,
+      "step": 2000
+    },
+    {
+      "epoch": 0.5414185165132648,
+      "eval_loss": 0.14294707775115967,
+      "eval_runtime": 103.5797,
+      "eval_samples_per_second": 4.943,
+      "eval_steps_per_second": 0.154,
+      "step": 2000
+    },
+    {
+      "epoch": 0.5441256090958311,
+      "grad_norm": 2.234375,
+      "learning_rate": 2.636192861615529e-05,
+      "loss": 0.099,
+      "step": 2010
+    },
+    {
+      "epoch": 0.5468327016783974,
+      "grad_norm": 1.2421875,
+      "learning_rate": 2.6205385097056983e-05,
+      "loss": 0.1048,
+      "step": 2020
+    },
+    {
+      "epoch": 0.5495397942609638,
+      "grad_norm": 1.984375,
+      "learning_rate": 2.604884157795867e-05,
+      "loss": 0.1117,
+      "step": 2030
+    },
+    {
+      "epoch": 0.5522468868435301,
+      "grad_norm": 1.6328125,
+      "learning_rate": 2.5892298058860364e-05,
+      "loss": 0.1291,
+      "step": 2040
+    },
+    {
+      "epoch": 0.5549539794260964,
+      "grad_norm": 1.4765625,
+      "learning_rate": 2.573575453976205e-05,
+      "loss": 0.1529,
+      "step": 2050
+    },
+    {
+      "epoch": 0.5576610720086627,
+      "grad_norm": 1.203125,
+      "learning_rate": 2.5579211020663746e-05,
+      "loss": 0.0559,
+      "step": 2060
+    },
+    {
+      "epoch": 0.5603681645912291,
+      "grad_norm": 1.40625,
+      "learning_rate": 2.5422667501565433e-05,
+      "loss": 0.0992,
+      "step": 2070
+    },
+    {
+      "epoch": 0.5630752571737954,
+      "grad_norm": 0.7578125,
+      "learning_rate": 2.5266123982467127e-05,
+      "loss": 0.0876,
+      "step": 2080
+    },
+    {
+      "epoch": 0.5657823497563617,
+      "grad_norm": 1.6328125,
+      "learning_rate": 2.510958046336882e-05,
+      "loss": 0.1442,
+      "step": 2090
+    },
+    {
+      "epoch": 0.568489442338928,
+      "grad_norm": 0.734375,
+      "learning_rate": 2.495303694427051e-05,
+      "loss": 0.073,
+      "step": 2100
+    },
+    {
+      "epoch": 0.568489442338928,
+      "eval_loss": 0.1371915191411972,
+      "eval_runtime": 102.7955,
+      "eval_samples_per_second": 4.981,
+      "eval_steps_per_second": 0.156,
+      "step": 2100
+    },
+    {
+      "epoch": 0.5711965349214944,
+      "grad_norm": 1.4453125,
+      "learning_rate": 2.47964934251722e-05,
+      "loss": 0.0818,
+      "step": 2110
+    },
+    {
+      "epoch": 0.5739036275040607,
+      "grad_norm": 1.0,
+      "learning_rate": 2.463994990607389e-05,
+      "loss": 0.0785,
+      "step": 2120
+    },
+    {
+      "epoch": 0.576610720086627,
+      "grad_norm": 2.09375,
+      "learning_rate": 2.448340638697558e-05,
+      "loss": 0.0966,
+      "step": 2130
+    },
+    {
+      "epoch": 0.5793178126691932,
+      "grad_norm": 1.3828125,
+      "learning_rate": 2.432686286787727e-05,
+      "loss": 0.0805,
+      "step": 2140
+    },
+    {
+      "epoch": 0.5820249052517596,
+      "grad_norm": 0.9375,
+      "learning_rate": 2.4170319348778962e-05,
+      "loss": 0.1032,
+      "step": 2150
+    },
+    {
+      "epoch": 0.584731997834326,
+      "grad_norm": 1.9765625,
+      "learning_rate": 2.4013775829680653e-05,
+      "loss": 0.1314,
+      "step": 2160
+    },
+    {
+      "epoch": 0.5874390904168922,
+      "grad_norm": 1.5078125,
+      "learning_rate": 2.3857232310582343e-05,
+      "loss": 0.1026,
+      "step": 2170
+    },
+    {
+      "epoch": 0.5901461829994585,
+      "grad_norm": 0.58984375,
+      "learning_rate": 2.3700688791484034e-05,
+      "loss": 0.0858,
+      "step": 2180
+    },
+    {
+      "epoch": 0.5928532755820249,
+      "grad_norm": 0.51171875,
+      "learning_rate": 2.3544145272385725e-05,
+      "loss": 0.069,
+      "step": 2190
+    },
+    {
+      "epoch": 0.5955603681645912,
+      "grad_norm": 0.94140625,
+      "learning_rate": 2.3387601753287412e-05,
+      "loss": 0.1238,
+      "step": 2200
+    },
+    {
+      "epoch": 0.5955603681645912,
+      "eval_loss": 0.1365373730659485,
+      "eval_runtime": 103.5823,
+      "eval_samples_per_second": 4.943,
+      "eval_steps_per_second": 0.154,
+      "step": 2200
+    },
+    {
+      "epoch": 0.5982674607471575,
+      "grad_norm": 1.0390625,
+      "learning_rate": 2.3231058234189106e-05,
+      "loss": 0.1196,
+      "step": 2210
+    },
+    {
+      "epoch": 0.6009745533297238,
+      "grad_norm": 1.1875,
+      "learning_rate": 2.3074514715090797e-05,
+      "loss": 0.111,
+      "step": 2220
+    },
+    {
+      "epoch": 0.6036816459122902,
+      "grad_norm": 0.5546875,
+      "learning_rate": 2.2917971195992488e-05,
+      "loss": 0.0875,
+      "step": 2230
+    },
+    {
+      "epoch": 0.6063887384948565,
+      "grad_norm": 1.2421875,
+      "learning_rate": 2.2761427676894178e-05,
+      "loss": 0.1029,
+      "step": 2240
+    },
+    {
+      "epoch": 0.6090958310774228,
+      "grad_norm": 1.3671875,
+      "learning_rate": 2.260488415779587e-05,
+      "loss": 0.0933,
+      "step": 2250
+    },
+    {
+      "epoch": 0.6118029236599891,
+      "grad_norm": 1.140625,
+      "learning_rate": 2.244834063869756e-05,
+      "loss": 0.1009,
+      "step": 2260
+    },
+    {
+      "epoch": 0.6145100162425555,
+      "grad_norm": 2.234375,
+      "learning_rate": 2.229179711959925e-05,
+      "loss": 0.135,
+      "step": 2270
+    },
+    {
+      "epoch": 0.6172171088251218,
+      "grad_norm": 1.671875,
+      "learning_rate": 2.213525360050094e-05,
+      "loss": 0.0773,
+      "step": 2280
+    },
+    {
+      "epoch": 0.6199242014076881,
+      "grad_norm": 0.9375,
+      "learning_rate": 2.1978710081402632e-05,
+      "loss": 0.0732,
+      "step": 2290
+    },
+    {
+      "epoch": 0.6226312939902545,
+      "grad_norm": 2.0625,
+      "learning_rate": 2.1822166562304323e-05,
+      "loss": 0.0851,
+      "step": 2300
+    },
+    {
+      "epoch": 0.6226312939902545,
+      "eval_loss": 0.1353287696838379,
+      "eval_runtime": 102.3815,
+      "eval_samples_per_second": 5.001,
+      "eval_steps_per_second": 0.156,
+      "step": 2300
+    },
+    {
+      "epoch": 0.6253383865728208,
+      "grad_norm": 0.71484375,
+      "learning_rate": 2.1665623043206013e-05,
+      "loss": 0.0633,
+      "step": 2310
+    },
+    {
+      "epoch": 0.6280454791553871,
+      "grad_norm": 1.328125,
+      "learning_rate": 2.1509079524107704e-05,
+      "loss": 0.0825,
+      "step": 2320
+    },
+    {
+      "epoch": 0.6307525717379534,
+      "grad_norm": 0.515625,
+      "learning_rate": 2.1352536005009395e-05,
+      "loss": 0.1016,
+      "step": 2330
+    },
+    {
+      "epoch": 0.6334596643205198,
+      "grad_norm": 2.40625,
+      "learning_rate": 2.1195992485911085e-05,
+      "loss": 0.0942,
+      "step": 2340
+    },
+    {
+      "epoch": 0.6361667569030861,
+      "grad_norm": 1.75,
+      "learning_rate": 2.1039448966812776e-05,
+      "loss": 0.086,
+      "step": 2350
+    },
+    {
+      "epoch": 0.6388738494856524,
+      "grad_norm": 1.328125,
+      "learning_rate": 2.0882905447714467e-05,
+      "loss": 0.1095,
+      "step": 2360
+    },
+    {
+      "epoch": 0.6415809420682187,
+      "grad_norm": 0.63671875,
+      "learning_rate": 2.0726361928616157e-05,
+      "loss": 0.0711,
+      "step": 2370
+    },
+    {
+      "epoch": 0.6442880346507851,
+      "grad_norm": 1.3515625,
+      "learning_rate": 2.0569818409517845e-05,
+      "loss": 0.0677,
+      "step": 2380
+    },
+    {
+      "epoch": 0.6469951272333514,
+      "grad_norm": 2.046875,
+      "learning_rate": 2.0413274890419535e-05,
+      "loss": 0.0892,
+      "step": 2390
+    },
+    {
+      "epoch": 0.6497022198159177,
+      "grad_norm": 1.640625,
+      "learning_rate": 2.0256731371321226e-05,
+      "loss": 0.0955,
+      "step": 2400
+    },
+    {
+      "epoch": 0.6497022198159177,
+      "eval_loss": 0.13519829511642456,
+      "eval_runtime": 121.253,
+      "eval_samples_per_second": 4.223,
+      "eval_steps_per_second": 0.132,
+      "step": 2400
+    },
+    {
+      "epoch": 0.652409312398484,
+      "grad_norm": 1.265625,
+      "learning_rate": 2.0100187852222917e-05,
+      "loss": 0.1012,
+      "step": 2410
+    },
+    {
+      "epoch": 0.6551164049810504,
+      "grad_norm": 0.984375,
+      "learning_rate": 1.9943644333124608e-05,
+      "loss": 0.1008,
+      "step": 2420
+    },
+    {
+      "epoch": 0.6578234975636167,
+      "grad_norm": 0.9453125,
+      "learning_rate": 1.9787100814026298e-05,
+      "loss": 0.1026,
+      "step": 2430
+    },
+    {
+      "epoch": 0.660530590146183,
+      "grad_norm": 0.95703125,
+      "learning_rate": 1.963055729492799e-05,
+      "loss": 0.1021,
+      "step": 2440
+    },
+    {
+      "epoch": 0.6632376827287493,
+      "grad_norm": 1.4375,
+      "learning_rate": 1.947401377582968e-05,
+      "loss": 0.1098,
+      "step": 2450
+    },
+    {
+      "epoch": 0.6659447753113157,
+      "grad_norm": 0.375,
+      "learning_rate": 1.931747025673137e-05,
+      "loss": 0.0828,
+      "step": 2460
+    },
+    {
+      "epoch": 0.668651867893882,
+      "grad_norm": 0.94140625,
+      "learning_rate": 1.9160926737633064e-05,
+      "loss": 0.0612,
+      "step": 2470
+    },
+    {
+      "epoch": 0.6713589604764483,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.9004383218534755e-05,
+      "loss": 0.1367,
+      "step": 2480
+    },
+    {
+      "epoch": 0.6740660530590146,
+      "grad_norm": 1.1796875,
+      "learning_rate": 1.8847839699436446e-05,
+      "loss": 0.0647,
+      "step": 2490
+    },
+    {
+      "epoch": 0.676773145641581,
+      "grad_norm": 0.5703125,
+      "learning_rate": 1.8691296180338137e-05,
+      "loss": 0.0973,
+      "step": 2500
+    },
+    {
+      "epoch": 0.676773145641581,
+      "eval_loss": 0.13495835661888123,
+      "eval_runtime": 105.5301,
+      "eval_samples_per_second": 4.852,
+      "eval_steps_per_second": 0.152,
+      "step": 2500
+    },
+    {
+      "epoch": 0.6794802382241473,
+      "grad_norm": 1.2890625,
+      "learning_rate": 1.8534752661239827e-05,
+      "loss": 0.0775,
+      "step": 2510
+    },
+    {
+      "epoch": 0.6821873308067136,
+      "grad_norm": 0.96484375,
+      "learning_rate": 1.8378209142141518e-05,
+      "loss": 0.0836,
+      "step": 2520
+    },
+    {
+      "epoch": 0.6848944233892799,
+      "grad_norm": 2.5625,
+      "learning_rate": 1.822166562304321e-05,
+      "loss": 0.1118,
+      "step": 2530
+    },
+    {
+      "epoch": 0.6876015159718463,
+      "grad_norm": 1.1328125,
+      "learning_rate": 1.80651221039449e-05,
+      "loss": 0.0889,
+      "step": 2540
+    },
+    {
+      "epoch": 0.6903086085544126,
+      "grad_norm": 0.71484375,
+      "learning_rate": 1.790857858484659e-05,
+      "loss": 0.0982,
+      "step": 2550
+    },
+    {
+      "epoch": 0.6930157011369789,
+      "grad_norm": 2.125,
+      "learning_rate": 1.775203506574828e-05,
+      "loss": 0.1131,
+      "step": 2560
+    },
+    {
+      "epoch": 0.6957227937195453,
+      "grad_norm": 1.015625,
+      "learning_rate": 1.7595491546649968e-05,
+      "loss": 0.1142,
+      "step": 2570
+    },
+    {
+      "epoch": 0.6984298863021116,
+      "grad_norm": 0.3671875,
+      "learning_rate": 1.743894802755166e-05,
+      "loss": 0.0911,
+      "step": 2580
+    },
+    {
+      "epoch": 0.7011369788846779,
+      "grad_norm": 1.2421875,
+      "learning_rate": 1.728240450845335e-05,
+      "loss": 0.0774,
+      "step": 2590
+    },
+    {
+      "epoch": 0.7038440714672441,
+      "grad_norm": 0.9453125,
+      "learning_rate": 1.712586098935504e-05,
+      "loss": 0.0933,
+      "step": 2600
+    },
+    {
+      "epoch": 0.7038440714672441,
+      "eval_loss": 0.13393962383270264,
+      "eval_runtime": 103.5576,
+      "eval_samples_per_second": 4.944,
+      "eval_steps_per_second": 0.155,
+      "step": 2600
+    },
+    {
+      "epoch": 0.7065511640498106,
+      "grad_norm": 1.234375,
+      "learning_rate": 1.696931747025673e-05,
+      "loss": 0.1073,
+      "step": 2610
+    },
+    {
+      "epoch": 0.7092582566323768,
+      "grad_norm": 2.203125,
+      "learning_rate": 1.681277395115842e-05,
+      "loss": 0.097,
+      "step": 2620
+    },
+    {
+      "epoch": 0.7119653492149431,
+      "grad_norm": 0.94140625,
+      "learning_rate": 1.6656230432060112e-05,
+      "loss": 0.0772,
+      "step": 2630
+    },
+    {
+      "epoch": 0.7146724417975094,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.6499686912961803e-05,
+      "loss": 0.0935,
+      "step": 2640
+    },
+    {
+      "epoch": 0.7173795343800758,
+      "grad_norm": 0.267578125,
+      "learning_rate": 1.6343143393863494e-05,
+      "loss": 0.0924,
+      "step": 2650
+    },
+    {
+      "epoch": 0.7200866269626421,
+      "grad_norm": 1.359375,
+      "learning_rate": 1.6186599874765184e-05,
+      "loss": 0.1057,
+      "step": 2660
+    },
+    {
+      "epoch": 0.7227937195452084,
+      "grad_norm": 1.4296875,
+      "learning_rate": 1.6030056355666875e-05,
+      "loss": 0.0865,
+      "step": 2670
+    },
+    {
+      "epoch": 0.7255008121277747,
+      "grad_norm": 1.09375,
+      "learning_rate": 1.5873512836568566e-05,
+      "loss": 0.0957,
+      "step": 2680
+    },
+    {
+      "epoch": 0.7282079047103411,
+      "grad_norm": 1.09375,
+      "learning_rate": 1.5716969317470257e-05,
+      "loss": 0.107,
+      "step": 2690
+    },
+    {
+      "epoch": 0.7309149972929074,
+      "grad_norm": 0.734375,
+      "learning_rate": 1.5560425798371947e-05,
+      "loss": 0.0976,
+      "step": 2700
+    },
+    {
+      "epoch": 0.7309149972929074,
+      "eval_loss": 0.13289867341518402,
+      "eval_runtime": 104.5232,
+      "eval_samples_per_second": 4.898,
+      "eval_steps_per_second": 0.153,
+      "step": 2700
+    },
+    {
+      "epoch": 0.7336220898754737,
+      "grad_norm": 0.34765625,
+      "learning_rate": 1.5403882279273638e-05,
+      "loss": 0.0918,
+      "step": 2710
+    },
+    {
+      "epoch": 0.73632918245804,
+      "grad_norm": 1.6875,
+      "learning_rate": 1.5247338760175329e-05,
+      "loss": 0.1113,
+      "step": 2720
+    },
+    {
+      "epoch": 0.7390362750406064,
+      "grad_norm": 1.328125,
+      "learning_rate": 1.5090795241077021e-05,
+      "loss": 0.1221,
+      "step": 2730
+    },
+    {
+      "epoch": 0.7417433676231727,
+      "grad_norm": 1.0390625,
+      "learning_rate": 1.4934251721978712e-05,
+      "loss": 0.1236,
+      "step": 2740
+    },
+    {
+      "epoch": 0.744450460205739,
+      "grad_norm": 0.95703125,
+      "learning_rate": 1.4777708202880403e-05,
+      "loss": 0.119,
+      "step": 2750
+    },
+    {
+      "epoch": 0.7471575527883053,
+      "grad_norm": 3.125,
+      "learning_rate": 1.4621164683782093e-05,
+      "loss": 0.1141,
+      "step": 2760
+    },
+    {
+      "epoch": 0.7498646453708717,
+      "grad_norm": 0.94921875,
+      "learning_rate": 1.4464621164683784e-05,
+      "loss": 0.106,
+      "step": 2770
+    },
+    {
+      "epoch": 0.752571737953438,
+      "grad_norm": 1.1640625,
+      "learning_rate": 1.4308077645585475e-05,
+      "loss": 0.0963,
+      "step": 2780
+    },
+    {
+      "epoch": 0.7552788305360043,
+      "grad_norm": 1.234375,
+      "learning_rate": 1.4151534126487165e-05,
+      "loss": 0.0946,
+      "step": 2790
+    },
+    {
+      "epoch": 0.7579859231185706,
+      "grad_norm": 1.71875,
+      "learning_rate": 1.3994990607388856e-05,
+      "loss": 0.0985,
+      "step": 2800
+    },
+    {
+      "epoch": 0.7579859231185706,
+      "eval_loss": 0.13054493069648743,
+      "eval_runtime": 102.656,
+      "eval_samples_per_second": 4.988,
+      "eval_steps_per_second": 0.156,
+      "step": 2800
+    },
+    {
+      "epoch": 0.760693015701137,
+      "grad_norm": 1.1875,
+      "learning_rate": 1.3838447088290547e-05,
+      "loss": 0.1329,
+      "step": 2810
+    },
+    {
+      "epoch": 0.7634001082837033,
+      "grad_norm": 2.21875,
+      "learning_rate": 1.3681903569192236e-05,
+      "loss": 0.1072,
+      "step": 2820
+    },
+    {
+      "epoch": 0.7661072008662696,
+      "grad_norm": 1.390625,
+      "learning_rate": 1.3525360050093926e-05,
+      "loss": 0.0953,
+      "step": 2830
+    },
+    {
+      "epoch": 0.7688142934488359,
+      "grad_norm": 1.046875,
+      "learning_rate": 1.3368816530995617e-05,
+      "loss": 0.0912,
+      "step": 2840
+    },
+    {
+      "epoch": 0.7715213860314023,
+      "grad_norm": 0.984375,
+      "learning_rate": 1.3212273011897308e-05,
+      "loss": 0.1132,
+      "step": 2850
+    },
+    {
+      "epoch": 0.7742284786139686,
+      "grad_norm": 0.5234375,
+      "learning_rate": 1.3055729492798999e-05,
+      "loss": 0.1059,
+      "step": 2860
+    },
+    {
+      "epoch": 0.7769355711965349,
+      "grad_norm": 0.99609375,
+      "learning_rate": 1.289918597370069e-05,
+      "loss": 0.1182,
+      "step": 2870
+    },
+    {
+      "epoch": 0.7796426637791013,
+      "grad_norm": 0.921875,
+      "learning_rate": 1.274264245460238e-05,
+      "loss": 0.0647,
+      "step": 2880
+    },
+    {
+      "epoch": 0.7823497563616676,
+      "grad_norm": 1.609375,
+      "learning_rate": 1.258609893550407e-05,
+      "loss": 0.1049,
+      "step": 2890
+    },
+    {
+      "epoch": 0.7850568489442339,
+      "grad_norm": 2.53125,
+      "learning_rate": 1.2429555416405761e-05,
+      "loss": 0.1223,
+      "step": 2900
+    },
+    {
+      "epoch": 0.7850568489442339,
+      "eval_loss": 0.13052764534950256,
+      "eval_runtime": 114.8255,
+      "eval_samples_per_second": 4.459,
+      "eval_steps_per_second": 0.139,
+      "step": 2900
+    },
+    {
+      "epoch": 0.7877639415268002,
+      "grad_norm": 1.59375,
+      "learning_rate": 1.2273011897307452e-05,
+      "loss": 0.0768,
+      "step": 2910
+    },
+    {
+      "epoch": 0.7904710341093666,
+      "grad_norm": 1.0,
+      "learning_rate": 1.2116468378209143e-05,
+      "loss": 0.0902,
+      "step": 2920
+    },
+    {
+      "epoch": 0.7931781266919329,
+      "grad_norm": 1.0546875,
+      "learning_rate": 1.1959924859110834e-05,
+      "loss": 0.1059,
+      "step": 2930
+    },
+    {
+      "epoch": 0.7958852192744992,
+      "grad_norm": 0.66796875,
+      "learning_rate": 1.1803381340012524e-05,
+      "loss": 0.0919,
+      "step": 2940
+    },
+    {
+      "epoch": 0.7985923118570655,
+      "grad_norm": 1.859375,
+      "learning_rate": 1.1646837820914215e-05,
+      "loss": 0.0779,
+      "step": 2950
+    },
+    {
+      "epoch": 0.8012994044396319,
+      "grad_norm": 1.1328125,
+      "learning_rate": 1.1490294301815906e-05,
+      "loss": 0.1019,
+      "step": 2960
+    },
+    {
+      "epoch": 0.8040064970221982,
+      "grad_norm": 0.8671875,
+      "learning_rate": 1.1333750782717596e-05,
+      "loss": 0.0909,
+      "step": 2970
+    },
+    {
+      "epoch": 0.8067135896047645,
+      "grad_norm": 0.435546875,
+      "learning_rate": 1.1177207263619287e-05,
+      "loss": 0.0973,
+      "step": 2980
+    },
+    {
+      "epoch": 0.8094206821873308,
+      "grad_norm": 1.453125,
+      "learning_rate": 1.1020663744520978e-05,
+      "loss": 0.0875,
+      "step": 2990
+    },
+    {
+      "epoch": 0.8121277747698972,
+      "grad_norm": 0.71484375,
+      "learning_rate": 1.0864120225422668e-05,
+      "loss": 0.1053,
+      "step": 3000
+    },
+    {
+      "epoch": 0.8121277747698972,
+      "eval_loss": 0.13060268759727478,
+      "eval_runtime": 108.9657,
+      "eval_samples_per_second": 4.699,
+      "eval_steps_per_second": 0.147,
+      "step": 3000
+    },
+    {
+      "epoch": 0.8148348673524635,
+      "grad_norm": 1.140625,
+      "learning_rate": 1.070757670632436e-05,
+      "loss": 0.0808,
+      "step": 3010
+    },
+    {
+      "epoch": 0.8175419599350298,
+      "grad_norm": 2.140625,
+      "learning_rate": 1.0551033187226048e-05,
+      "loss": 0.0766,
+      "step": 3020
+    },
+    {
+      "epoch": 0.8202490525175961,
+      "grad_norm": 1.0546875,
+      "learning_rate": 1.0394489668127739e-05,
+      "loss": 0.073,
+      "step": 3030
+    },
+    {
+      "epoch": 0.8229561451001625,
+      "grad_norm": 1.859375,
+      "learning_rate": 1.023794614902943e-05,
+      "loss": 0.108,
+      "step": 3040
+    },
+    {
+      "epoch": 0.8256632376827288,
+      "grad_norm": 0.69140625,
+      "learning_rate": 1.008140262993112e-05,
+      "loss": 0.0815,
+      "step": 3050
+    },
+    {
+      "epoch": 0.828370330265295,
+      "grad_norm": 0.80859375,
+      "learning_rate": 9.924859110832813e-06,
+      "loss": 0.0965,
+      "step": 3060
+    },
+    {
+      "epoch": 0.8310774228478613,
+      "grad_norm": 1.953125,
+      "learning_rate": 9.768315591734503e-06,
+      "loss": 0.1145,
+      "step": 3070
+    },
+    {
+      "epoch": 0.8337845154304278,
+      "grad_norm": 0.734375,
+      "learning_rate": 9.611772072636194e-06,
+      "loss": 0.1319,
+      "step": 3080
+    },
+    {
+      "epoch": 0.836491608012994,
+      "grad_norm": 1.671875,
+      "learning_rate": 9.455228553537885e-06,
+      "loss": 0.0871,
+      "step": 3090
+    },
+    {
+      "epoch": 0.8391987005955603,
+      "grad_norm": 1.515625,
+      "learning_rate": 9.298685034439576e-06,
+      "loss": 0.1207,
+      "step": 3100
+    },
+    {
+      "epoch": 0.8391987005955603,
+      "eval_loss": 0.1296384036540985,
+      "eval_runtime": 103.7944,
+      "eval_samples_per_second": 4.933,
+      "eval_steps_per_second": 0.154,
+      "step": 3100
+    },
+    {
+      "epoch": 0.8419057931781266,
+      "grad_norm": 2.203125,
+      "learning_rate": 9.142141515341266e-06,
+      "loss": 0.1177,
+      "step": 3110
+    },
+    {
+      "epoch": 0.844612885760693,
+      "grad_norm": 1.1171875,
+      "learning_rate": 8.985597996242955e-06,
+      "loss": 0.1111,
+      "step": 3120
+    },
+    {
+      "epoch": 0.8473199783432593,
+      "grad_norm": 0.384765625,
+      "learning_rate": 8.829054477144646e-06,
+      "loss": 0.0774,
+      "step": 3130
+    },
+    {
+      "epoch": 0.8500270709258256,
+      "grad_norm": 1.3984375,
+      "learning_rate": 8.672510958046337e-06,
+      "loss": 0.0864,
+      "step": 3140
+    },
+    {
+      "epoch": 0.852734163508392,
+      "grad_norm": 0.42578125,
+      "learning_rate": 8.515967438948027e-06,
+      "loss": 0.0914,
+      "step": 3150
+    },
+    {
+      "epoch": 0.8554412560909583,
+      "grad_norm": 0.453125,
+      "learning_rate": 8.359423919849718e-06,
+      "loss": 0.0947,
+      "step": 3160
+    },
+    {
+      "epoch": 0.8581483486735246,
+      "grad_norm": 1.265625,
+      "learning_rate": 8.202880400751409e-06,
+      "loss": 0.0812,
+      "step": 3170
+    },
+    {
+      "epoch": 0.8608554412560909,
+      "grad_norm": 1.296875,
+      "learning_rate": 8.0463368816531e-06,
+      "loss": 0.0924,
+      "step": 3180
+    },
+    {
+      "epoch": 0.8635625338386573,
+      "grad_norm": 0.69921875,
+      "learning_rate": 7.889793362554792e-06,
+      "loss": 0.1047,
+      "step": 3190
+    },
+    {
+      "epoch": 0.8662696264212236,
+      "grad_norm": 2.0,
+      "learning_rate": 7.733249843456483e-06,
+      "loss": 0.1071,
+      "step": 3200
+    },
+    {
+      "epoch": 0.8662696264212236,
+      "eval_loss": 0.128614142537117,
+      "eval_runtime": 105.0877,
+      "eval_samples_per_second": 4.872,
+      "eval_steps_per_second": 0.152,
+      "step": 3200
+    },
+    {
+      "epoch": 0.8689767190037899,
+      "grad_norm": 1.8984375,
+      "learning_rate": 7.576706324358172e-06,
+      "loss": 0.0834,
+      "step": 3210
+    },
+    {
+      "epoch": 0.8716838115863562,
+      "grad_norm": 0.94921875,
+      "learning_rate": 7.420162805259863e-06,
+      "loss": 0.1098,
+      "step": 3220
+    },
+    {
+      "epoch": 0.8743909041689226,
+      "grad_norm": 1.2890625,
+      "learning_rate": 7.263619286161554e-06,
+      "loss": 0.1152,
+      "step": 3230
+    },
+    {
+      "epoch": 0.8770979967514889,
+      "grad_norm": 0.63671875,
+      "learning_rate": 7.107075767063244e-06,
+      "loss": 0.1093,
+      "step": 3240
+    },
+    {
+      "epoch": 0.8798050893340552,
+      "grad_norm": 0.93359375,
+      "learning_rate": 6.950532247964934e-06,
+      "loss": 0.1116,
+      "step": 3250
+    },
+    {
+      "epoch": 0.8825121819166215,
+      "grad_norm": 1.421875,
+      "learning_rate": 6.793988728866625e-06,
+      "loss": 0.0824,
+      "step": 3260
+    },
+    {
+      "epoch": 0.8852192744991879,
+      "grad_norm": 1.6953125,
+      "learning_rate": 6.637445209768316e-06,
+      "loss": 0.1241,
+      "step": 3270
+    },
+    {
+      "epoch": 0.8879263670817542,
+      "grad_norm": 1.0859375,
+      "learning_rate": 6.4809016906700065e-06,
+      "loss": 0.0967,
+      "step": 3280
+    },
+    {
+      "epoch": 0.8906334596643205,
+      "grad_norm": 1.125,
+      "learning_rate": 6.324358171571697e-06,
+      "loss": 0.0893,
+      "step": 3290
+    },
+    {
+      "epoch": 0.8933405522468868,
+      "grad_norm": 1.1875,
+      "learning_rate": 6.167814652473388e-06,
+      "loss": 0.1081,
+      "step": 3300
+    },
+    {
+      "epoch": 0.8933405522468868,
+      "eval_loss": 0.12801626324653625,
+      "eval_runtime": 103.0737,
+      "eval_samples_per_second": 4.967,
+      "eval_steps_per_second": 0.155,
+      "step": 3300
+    },
+    {
+      "epoch": 0.8960476448294532,
+      "grad_norm": 1.3984375,
+      "learning_rate": 6.011271133375079e-06,
+      "loss": 0.1199,
+      "step": 3310
+    },
+    {
+      "epoch": 0.8987547374120195,
+      "grad_norm": 1.640625,
+      "learning_rate": 5.854727614276769e-06,
+      "loss": 0.0997,
+      "step": 3320
+    },
+    {
+      "epoch": 0.9014618299945858,
+      "grad_norm": 0.921875,
+      "learning_rate": 5.69818409517846e-06,
+      "loss": 0.1118,
+      "step": 3330
+    },
+    {
+      "epoch": 0.9041689225771521,
+      "grad_norm": 1.3828125,
+      "learning_rate": 5.54164057608015e-06,
+      "loss": 0.083,
+      "step": 3340
+    },
+    {
+      "epoch": 0.9068760151597185,
+      "grad_norm": 1.609375,
+      "learning_rate": 5.3850970569818414e-06,
+      "loss": 0.0765,
+      "step": 3350
+    },
+    {
+      "epoch": 0.9095831077422848,
+      "grad_norm": 1.0546875,
+      "learning_rate": 5.228553537883532e-06,
+      "loss": 0.0925,
+      "step": 3360
+    },
+    {
+      "epoch": 0.9122902003248511,
+      "grad_norm": 1.203125,
+      "learning_rate": 5.072010018785223e-06,
+      "loss": 0.0941,
+      "step": 3370
+    },
+    {
+      "epoch": 0.9149972929074174,
+      "grad_norm": 1.1796875,
+      "learning_rate": 4.9154664996869136e-06,
+      "loss": 0.1193,
+      "step": 3380
+    },
+    {
+      "epoch": 0.9177043854899838,
+      "grad_norm": 0.333984375,
+      "learning_rate": 4.758922980588603e-06,
+      "loss": 0.1124,
+      "step": 3390
+    },
+    {
+      "epoch": 0.9204114780725501,
+      "grad_norm": 1.6796875,
+      "learning_rate": 4.602379461490294e-06,
+      "loss": 0.0924,
+      "step": 3400
+    },
+    {
+      "epoch": 0.9204114780725501,
+      "eval_loss": 0.12878485023975372,
+      "eval_runtime": 123.0744,
+      "eval_samples_per_second": 4.16,
+      "eval_steps_per_second": 0.13,
+      "step": 3400
+    },
+    {
+      "epoch": 0.9231185706551164,
+      "grad_norm": 0.86328125,
+      "learning_rate": 4.445835942391985e-06,
+      "loss": 0.0938,
+      "step": 3410
+    },
+    {
+      "epoch": 0.9258256632376828,
+      "grad_norm": 2.328125,
+      "learning_rate": 4.289292423293676e-06,
+      "loss": 0.0839,
+      "step": 3420
+    },
+    {
+      "epoch": 0.9285327558202491,
+      "grad_norm": 1.7578125,
+      "learning_rate": 4.132748904195367e-06,
+      "loss": 0.1368,
+      "step": 3430
+    },
+    {
+      "epoch": 0.9312398484028154,
+      "grad_norm": 2.515625,
+      "learning_rate": 3.976205385097057e-06,
+      "loss": 0.1154,
+      "step": 3440
+    },
+    {
+      "epoch": 0.9339469409853817,
+      "grad_norm": 0.205078125,
+      "learning_rate": 3.819661865998748e-06,
+      "loss": 0.095,
+      "step": 3450
+    },
+    {
+      "epoch": 0.9366540335679481,
+      "grad_norm": 0.87890625,
+      "learning_rate": 3.6631183469004384e-06,
+      "loss": 0.0872,
+      "step": 3460
+    },
+    {
+      "epoch": 0.9393611261505144,
+      "grad_norm": 2.515625,
+      "learning_rate": 3.506574827802129e-06,
+      "loss": 0.092,
+      "step": 3470
+    },
+    {
+      "epoch": 0.9420682187330807,
+      "grad_norm": 1.125,
+      "learning_rate": 3.35003130870382e-06,
+      "loss": 0.0944,
+      "step": 3480
+    },
+    {
+      "epoch": 0.944775311315647,
+      "grad_norm": 1.578125,
+      "learning_rate": 3.193487789605511e-06,
+      "loss": 0.1366,
+      "step": 3490
+    },
+    {
+      "epoch": 0.9474824038982134,
+      "grad_norm": 0.5859375,
+      "learning_rate": 3.036944270507201e-06,
+      "loss": 0.0948,
+      "step": 3500
+    },
+    {
+      "epoch": 0.9474824038982134,
+      "eval_loss": 0.1291881799697876,
+      "eval_runtime": 104.6704,
+      "eval_samples_per_second": 4.892,
+      "eval_steps_per_second": 0.153,
+      "step": 3500
+    },
+    {
+      "epoch": 0.9501894964807797,
+      "grad_norm": 1.375,
+      "learning_rate": 2.880400751408892e-06,
+      "loss": 0.1181,
+      "step": 3510
+    },
+    {
+      "epoch": 0.952896589063346,
+      "grad_norm": 0.84375,
+      "learning_rate": 2.7238572323105826e-06,
+      "loss": 0.0819,
+      "step": 3520
+    },
+    {
+      "epoch": 0.9556036816459123,
+      "grad_norm": 0.9765625,
+      "learning_rate": 2.5673137132122733e-06,
+      "loss": 0.1134,
+      "step": 3530
+    },
+    {
+      "epoch": 0.9583107742284787,
+      "grad_norm": 0.5625,
+      "learning_rate": 2.410770194113964e-06,
+      "loss": 0.1183,
+      "step": 3540
+    },
+    {
+      "epoch": 0.961017866811045,
+      "grad_norm": 0.84765625,
+      "learning_rate": 2.2542266750156543e-06,
+      "loss": 0.1021,
+      "step": 3550
+    },
+    {
+      "epoch": 0.9637249593936112,
+      "grad_norm": 1.0390625,
+      "learning_rate": 2.0976831559173454e-06,
+      "loss": 0.0871,
+      "step": 3560
+    },
+    {
+      "epoch": 0.9664320519761775,
+      "grad_norm": 1.1171875,
+      "learning_rate": 1.9411396368190357e-06,
+      "loss": 0.1036,
+      "step": 3570
+    },
+    {
+      "epoch": 0.969139144558744,
+      "grad_norm": 0.546875,
+      "learning_rate": 1.7845961177207264e-06,
+      "loss": 0.0874,
+      "step": 3580
+    },
+    {
+      "epoch": 0.9718462371413102,
+      "grad_norm": 0.28515625,
+      "learning_rate": 1.6280525986224169e-06,
+      "loss": 0.085,
+      "step": 3590
+    },
+    {
+      "epoch": 0.9745533297238765,
+      "grad_norm": 0.90234375,
+      "learning_rate": 1.4715090795241078e-06,
+      "loss": 0.1162,
+      "step": 3600
+    },
+    {
+      "epoch": 0.9745533297238765,
+      "eval_loss": 0.12912487983703613,
+      "eval_runtime": 106.4924,
+      "eval_samples_per_second": 4.808,
+      "eval_steps_per_second": 0.15,
+      "step": 3600
+    },
+    {
+      "epoch": 0.9772604223064428,
+      "grad_norm": 0.74609375,
+      "learning_rate": 1.3149655604257985e-06,
+      "loss": 0.1066,
+      "step": 3610
+    },
+    {
+      "epoch": 0.9799675148890092,
+      "grad_norm": 1.4609375,
+      "learning_rate": 1.1584220413274892e-06,
+      "loss": 0.1533,
+      "step": 3620
+    },
+    {
+      "epoch": 0.9826746074715755,
+      "grad_norm": 0.68359375,
+      "learning_rate": 1.0018785222291797e-06,
+      "loss": 0.0717,
+      "step": 3630
+    },
+    {
+      "epoch": 0.9853817000541418,
+      "grad_norm": 0.94921875,
+      "learning_rate": 8.453350031308704e-07,
+      "loss": 0.0667,
+      "step": 3640
+    },
+    {
+      "epoch": 0.9880887926367081,
+      "grad_norm": 0.427734375,
+      "learning_rate": 6.887914840325611e-07,
+      "loss": 0.1175,
+      "step": 3650
+    },
+    {
+      "epoch": 0.9907958852192745,
+      "grad_norm": 1.046875,
+      "learning_rate": 5.322479649342517e-07,
+      "loss": 0.0871,
+      "step": 3660
+    },
+    {
+      "epoch": 0.9935029778018408,
+      "grad_norm": 1.8515625,
+      "learning_rate": 3.757044458359424e-07,
+      "loss": 0.122,
+      "step": 3670
+    },
+    {
+      "epoch": 0.9962100703844071,
+      "grad_norm": 0.77734375,
+      "learning_rate": 2.1916092673763307e-07,
+      "loss": 0.0922,
+      "step": 3680
+    },
+    {
+      "epoch": 0.9989171629669734,
+      "grad_norm": 3.140625,
+      "learning_rate": 6.261740763932373e-08,
+      "loss": 0.1331,
+      "step": 3690
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 3694,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "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": 1.0990508415397056e+18,
+  "train_batch_size": 32,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-3694/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:685ee2596072507a9c0ce1ae768b36fc3337189cbe14f6c946fc5fb838a3c04b
+size 5176

git_hash.txt

@@ -0,0 +1 @@
+a5b4a24649e45a6570b33cc01acb57807e4dfa5b

preprocessor_config.json

@@ -0,0 +1,29 @@
+{
+  "do_convert_rgb": true,
+  "do_normalize": true,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.48145466,
+    0.4578275,
+    0.40821073
+  ],
+  "image_processor_type": "Qwen2VLImageProcessor",
+  "image_std": [
+    0.26862954,
+    0.26130258,
+    0.27577711
+  ],
+  "max_pixels": 12845056,
+  "merge_size": 2,
+  "min_pixels": 3136,
+  "patch_size": 14,
+  "processor_class": "ColQwen2Processor",
+  "resample": 3,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "max_pixels": 12845056,
+    "min_pixels": 3136
+  },
+  "temporal_patch_size": 2
+}

results.json

@@ -0,0 +1 @@
+{"validation_set": {"ndcg_at_1": 0.81641, "ndcg_at_3": 0.87151, "ndcg_at_5": 0.87924, "ndcg_at_10": 0.88352, "ndcg_at_20": 0.88599, "ndcg_at_100": 0.89228, "ndcg_at_1000": 0.8945, "map_at_1": 0.81641, "map_at_3": 0.85872, "map_at_5": 0.86283, "map_at_10": 0.86451, "map_at_20": 0.86519, "map_at_100": 0.86613, "map_at_1000": 0.86625, "recall_at_1": 0.81641, "recall_at_3": 0.9082, "recall_at_5": 0.92773, "recall_at_10": 0.94141, "recall_at_20": 0.95117, "recall_at_100": 0.98438, "recall_at_1000": 1.0, "precision_at_1": 0.81641, "precision_at_3": 0.30273, "precision_at_5": 0.18555, "precision_at_10": 0.09414, "precision_at_20": 0.04756, "precision_at_100": 0.00984, "precision_at_1000": 0.001, "mrr_at_1": 0.8203125, "mrr_at_3": 0.8603515624999998, "mrr_at_5": 0.8628906249999998, "mrr_at_10": 0.8655544704861109, "mrr_at_20": 0.8665781348642675, "mrr_at_100": 0.8675737791279903, "mrr_at_1000": 0.8676711773156423, "naucs_at_1_max": 0.34667870891415986, "naucs_at_1_std": -0.20661616231925453, "naucs_at_1_diff1": 0.9396571084820124, "naucs_at_3_max": 0.5166729885733626, "naucs_at_3_std": 0.05950323676330907, "naucs_at_3_diff1": 0.9342216236825106, "naucs_at_5_max": 0.6329364969562152, "naucs_at_5_std": 0.2619697444169589, "naucs_at_5_diff1": 0.9376321560083725, "naucs_at_10_max": 0.660722672615821, "naucs_at_10_std": 0.437248626444334, "naucs_at_10_diff1": 0.9546510563783107, "naucs_at_20_max": 0.6658122614005673, "naucs_at_20_std": 0.4847758170885219, "naucs_at_20_diff1": 0.9508077573821819, "naucs_at_100_max": 0.6851951655323608, "naucs_at_100_std": 0.8932409444836575, "naucs_at_100_diff1": 0.9510016587980088, "naucs_at_1000_max": 1.0, "naucs_at_1000_std": 1.0, "naucs_at_1000_diff1": 1.0}, "syntheticDocQA_energy": {"ndcg_at_1": 0.92, "ndcg_at_3": 0.94762, "ndcg_at_5": 0.94762, "ndcg_at_10": 0.94762, "ndcg_at_20": 0.95512, "ndcg_at_100": 0.95512, "ndcg_at_1000": 0.95512, "map_at_1": 0.92, "map_at_3": 0.94, "map_at_5": 0.94, "map_at_10": 0.94, "map_at_20": 0.94201, "map_at_100": 0.94201, "map_at_1000": 0.94201, "recall_at_1": 0.92, "recall_at_3": 0.97, "recall_at_5": 0.97, "recall_at_10": 0.97, "recall_at_20": 1.0, "recall_at_100": 1.0, "recall_at_1000": 1.0, "precision_at_1": 0.92, "precision_at_3": 0.32333, "precision_at_5": 0.194, "precision_at_10": 0.097, "precision_at_20": 0.05, "precision_at_100": 0.01, "precision_at_1000": 0.001, "mrr_at_1": 0.92, "mrr_at_3": 0.94, "mrr_at_5": 0.94, "mrr_at_10": 0.94, "mrr_at_20": 0.9421501831501832, "mrr_at_100": 0.9421501831501832, "mrr_at_1000": 0.9421501831501832, "naucs_at_1_max": 0.5697362278244628, "naucs_at_1_std": -0.5218253968253984, "naucs_at_1_diff1": 0.9489379084967322, "naucs_at_3_max": 0.7642390289449176, "naucs_at_3_std": -0.8249299719887898, "naucs_at_3_diff1": 1.0, "naucs_at_5_max": 0.7642390289449118, "naucs_at_5_std": -0.8249299719887969, "naucs_at_5_diff1": 1.0, "naucs_at_10_max": 0.7642390289449118, "naucs_at_10_std": -0.8249299719887969, "naucs_at_10_diff1": 1.0, "naucs_at_20_max": 1.0, "naucs_at_20_std": 1.0, "naucs_at_20_diff1": 1.0, "naucs_at_100_max": NaN, "naucs_at_100_std": NaN, "naucs_at_100_diff1": NaN, "naucs_at_1000_max": NaN, "naucs_at_1000_std": NaN, "naucs_at_1000_diff1": NaN}, "syntheticDocQA_healthcare_industry": {"ndcg_at_1": 0.93, "ndcg_at_3": 0.97286, "ndcg_at_5": 0.97286, "ndcg_at_10": 0.97286, "ndcg_at_20": 0.97286, "ndcg_at_100": 0.97286, "ndcg_at_1000": 0.97286, "map_at_1": 0.93, "map_at_3": 0.96333, "map_at_5": 0.96333, "map_at_10": 0.96333, "map_at_20": 0.96333, "map_at_100": 0.96333, "map_at_1000": 0.96333, "recall_at_1": 0.93, "recall_at_3": 1.0, "recall_at_5": 1.0, "recall_at_10": 1.0, "recall_at_20": 1.0, "recall_at_100": 1.0, "recall_at_1000": 1.0, "precision_at_1": 0.93, "precision_at_3": 0.33333, "precision_at_5": 0.2, "precision_at_10": 0.1, "precision_at_20": 0.05, "precision_at_100": 0.01, "precision_at_1000": 0.001, "mrr_at_1": 0.93, "mrr_at_3": 0.9633333333333334, "mrr_at_5": 0.9633333333333334, "mrr_at_10": 0.9633333333333334, "mrr_at_20": 0.9633333333333334, "mrr_at_100": 0.9633333333333334, "mrr_at_1000": 0.9633333333333334, "naucs_at_1_max": 0.7465652927837806, "naucs_at_1_std": 0.10897692410297374, "naucs_at_1_diff1": 0.9626517273576113, "naucs_at_3_max": 1.0, "naucs_at_3_std": 1.0, "naucs_at_3_diff1": 1.0, "naucs_at_5_max": 1.0, "naucs_at_5_std": 1.0, "naucs_at_5_diff1": 1.0, "naucs_at_10_max": 1.0, "naucs_at_10_std": 1.0, "naucs_at_10_diff1": 1.0, "naucs_at_20_max": 1.0, "naucs_at_20_std": 1.0, "naucs_at_20_diff1": 1.0, "naucs_at_100_max": NaN, "naucs_at_100_std": NaN, "naucs_at_100_diff1": NaN, "naucs_at_1000_max": NaN, "naucs_at_1000_std": NaN, "naucs_at_1000_diff1": NaN}, "syntheticDocQA_artificial_intelligence_test": {"ndcg_at_1": 0.97, "ndcg_at_3": 0.98762, "ndcg_at_5": 0.98762, "ndcg_at_10": 0.98762, "ndcg_at_20": 0.98762, "ndcg_at_100": 0.98762, "ndcg_at_1000": 0.98762, "map_at_1": 0.97, "map_at_3": 0.98333, "map_at_5": 0.98333, "map_at_10": 0.98333, "map_at_20": 0.98333, "map_at_100": 0.98333, "map_at_1000": 0.98333, "recall_at_1": 0.97, "recall_at_3": 1.0, "recall_at_5": 1.0, "recall_at_10": 1.0, "recall_at_20": 1.0, "recall_at_100": 1.0, "recall_at_1000": 1.0, "precision_at_1": 0.97, "precision_at_3": 0.33333, "precision_at_5": 0.2, "precision_at_10": 0.1, "precision_at_20": 0.05, "precision_at_100": 0.01, "precision_at_1000": 0.001, "mrr_at_1": 0.97, "mrr_at_3": 0.9833333333333333, "mrr_at_5": 0.9833333333333333, "mrr_at_10": 0.9833333333333333, "mrr_at_20": 0.9833333333333333, "mrr_at_100": 0.9833333333333333, "mrr_at_1000": 0.9833333333333333, "naucs_at_1_max": 0.8638344226579567, "naucs_at_1_std": -0.28197945845004707, "naucs_at_1_diff1": 0.9564270152505465, "naucs_at_3_max": 1.0, "naucs_at_3_std": 1.0, "naucs_at_3_diff1": 1.0, "naucs_at_5_max": 1.0, "naucs_at_5_std": 1.0, "naucs_at_5_diff1": 1.0, "naucs_at_10_max": 1.0, "naucs_at_10_std": 1.0, "naucs_at_10_diff1": 1.0, "naucs_at_20_max": 1.0, "naucs_at_20_std": 1.0, "naucs_at_20_diff1": 1.0, "naucs_at_100_max": NaN, "naucs_at_100_std": NaN, "naucs_at_100_diff1": NaN, "naucs_at_1000_max": NaN, "naucs_at_1000_std": NaN, "naucs_at_1000_diff1": NaN}, "syntheticDocQA_government_reports": {"ndcg_at_1": 0.85, "ndcg_at_3": 0.92809, "ndcg_at_5": 0.9324, "ndcg_at_10": 0.9324, "ndcg_at_20": 0.93503, "ndcg_at_100": 0.93503, "ndcg_at_1000": 0.93503, "map_at_1": 0.85, "map_at_3": 0.91, "map_at_5": 0.9125, "map_at_10": 0.9125, "map_at_20": 0.91327, "map_at_100": 0.91327, "map_at_1000": 0.91327, "recall_at_1": 0.85, "recall_at_3": 0.98, "recall_at_5": 0.99, "recall_at_10": 0.99, "recall_at_20": 1.0, "recall_at_100": 1.0, "recall_at_1000": 1.0, "precision_at_1": 0.85, "precision_at_3": 0.32667, "precision_at_5": 0.198, "precision_at_10": 0.099, "precision_at_20": 0.05, "precision_at_100": 0.01, "precision_at_1000": 0.001, "mrr_at_1": 0.87, "mrr_at_3": 0.9266666666666667, "mrr_at_5": 0.9266666666666667, "mrr_at_10": 0.9266666666666667, "mrr_at_20": 0.9274358974358975, "mrr_at_100": 0.9274358974358975, "mrr_at_1000": 0.9274358974358975, "naucs_at_1_max": 0.47617962902700983, "naucs_at_1_std": 0.33045883501464407, "naucs_at_1_diff1": 0.8781646599414249, "naucs_at_3_max": 1.0, "naucs_at_3_std": 0.6381886087768379, "naucs_at_3_diff1": 1.0, "naucs_at_5_max": 1.0, "naucs_at_5_std": 0.7222222222222276, "naucs_at_5_diff1": 1.0, "naucs_at_10_max": 1.0, "naucs_at_10_std": 0.7222222222222276, "naucs_at_10_diff1": 1.0, "naucs_at_20_max": 1.0, "naucs_at_20_std": 1.0, "naucs_at_20_diff1": 1.0, "naucs_at_100_max": NaN, "naucs_at_100_std": NaN, "naucs_at_100_diff1": NaN, "naucs_at_1000_max": NaN, "naucs_at_1000_std": NaN, "naucs_at_1000_diff1": NaN}, "infovqa_subsampled": {"ndcg_at_1": 0.86, "ndcg_at_3": 0.88671, "ndcg_at_5": 0.89825, "ndcg_at_10": 0.90151, "ndcg_at_20": 0.90849, "ndcg_at_100": 0.91162, "ndcg_at_1000": 0.91339, "map_at_1": 0.86, "map_at_3": 0.88067, "map_at_5": 0.88707, "map_at_10": 0.88843, "map_at_20": 0.89056, "map_at_100": 0.89092, "map_at_1000": 0.89099, "recall_at_1": 0.86, "recall_at_3": 0.904, "recall_at_5": 0.932, "recall_at_10": 0.942, "recall_at_20": 0.968, "recall_at_100": 0.986, "recall_at_1000": 1.0, "precision_at_1": 0.86, "precision_at_3": 0.30133, "precision_at_5": 0.1864, "precision_at_10": 0.0942, "precision_at_20": 0.0484, "precision_at_100": 0.00986, "precision_at_1000": 0.001, "mrr_at_1": 0.858, "mrr_at_3": 0.8799999999999999, "mrr_at_5": 0.8859999999999999, "mrr_at_10": 0.8877992063492062, "mrr_at_20": 0.8897055999555997, "mrr_at_100": 0.8900658496714725, "mrr_at_1000": 0.8901272220832652, "naucs_at_1_max": 0.6271096023278367, "naucs_at_1_std": -0.043875571567133, "naucs_at_1_diff1": 0.9370583344880139, "naucs_at_3_max": 0.6881127450980395, "naucs_at_3_std": -0.021095938375348355, "naucs_at_3_diff1": 0.9083313881108002, "naucs_at_5_max": 0.8094002306805071, "naucs_at_5_std": 0.05795847750865092, "naucs_at_5_diff1": 0.9197424067666282, "naucs_at_10_max": 0.877426832802087, "naucs_at_10_std": 0.2035963810811664, "naucs_at_10_diff1": 0.9250619788145158, "naucs_at_20_max": 0.9557948179271724, "naucs_at_20_std": 0.38013538748832426, "naucs_at_20_diff1": 0.9325980392156829, "naucs_at_100_max": 0.9626517273576126, "naucs_at_100_std": 0.448312658396679, "naucs_at_100_diff1": 0.981325863678799, "naucs_at_1000_max": 1.0, "naucs_at_1000_std": 1.0, "naucs_at_1000_diff1": 1.0}, "docvqa_subsampled": {"ndcg_at_1": 0.47, "ndcg_at_3": 0.53112, "ndcg_at_5": 0.56176, "ndcg_at_10": 0.58036, "ndcg_at_20": 0.59124, "ndcg_at_100": 0.61372, "ndcg_at_1000": 0.62957, "map_at_1": 0.47, "map_at_3": 0.51567, "map_at_5": 0.53277, "map_at_10": 0.54035, "map_at_20": 0.54347, "map_at_100": 0.54625, "map_at_1000": 0.54684, "recall_at_1": 0.47, "recall_at_3": 0.576, "recall_at_5": 0.65, "recall_at_10": 0.708, "recall_at_20": 0.75, "recall_at_100": 0.876, "recall_at_1000": 1.0, "precision_at_1": 0.47, "precision_at_3": 0.192, "precision_at_5": 0.13, "precision_at_10": 0.0708, "precision_at_20": 0.0375, "precision_at_100": 0.00876, "precision_at_1000": 0.001, "mrr_at_1": 0.47, "mrr_at_3": 0.5140000000000002, "mrr_at_5": 0.5322999999999999, "mrr_at_10": 0.5393452380952379, "mrr_at_20": 0.5430581398134026, "mrr_at_100": 0.5455877170766977, "mrr_at_1000": 0.5461605124926291, "naucs_at_1_max": 0.5899734964859769, "naucs_at_1_std": 0.37540315668719265, "naucs_at_1_diff1": 0.8310658104669103, "naucs_at_3_max": 0.5424525576127407, "naucs_at_3_std": 0.425733305596006, "naucs_at_3_diff1": 0.7247359175276799, "naucs_at_5_max": 0.5197754089022696, "naucs_at_5_std": 0.5280852795182682, "naucs_at_5_diff1": 0.6994124827081123, "naucs_at_10_max": 0.4841097869266883, "naucs_at_10_std": 0.5795468783396347, "naucs_at_10_diff1": 0.6604425085310391, "naucs_at_20_max": 0.4785557299843014, "naucs_at_20_std": 0.6286844583987442, "naucs_at_20_diff1": 0.6366300366300363, "naucs_at_100_max": 0.531426717344952, "naucs_at_100_std": 0.8670699298538366, "naucs_at_100_diff1": 0.5190043570965052, "naucs_at_1000_max": 1.0, "naucs_at_1000_std": 1.0, "naucs_at_1000_diff1": 1.0}, "arxivqa_subsampled": {"ndcg_at_1": 0.8, "ndcg_at_3": 0.84812, "ndcg_at_5": 0.86369, "ndcg_at_10": 0.87151, "ndcg_at_20": 0.87714, "ndcg_at_100": 0.88221, "ndcg_at_1000": 0.88366, "map_at_1": 0.8, "map_at_3": 0.837, "map_at_5": 0.8456, "map_at_10": 0.84885, "map_at_20": 0.85044, "map_at_100": 0.85123, "map_at_1000": 0.85132, "recall_at_1": 0.8, "recall_at_3": 0.88, "recall_at_5": 0.918, "recall_at_10": 0.942, "recall_at_20": 0.964, "recall_at_100": 0.99, "recall_at_1000": 1.0, "precision_at_1": 0.8, "precision_at_3": 0.29333, "precision_at_5": 0.1836, "precision_at_10": 0.0942, "precision_at_20": 0.0482, "precision_at_100": 0.0099, "precision_at_1000": 0.001, "mrr_at_1": 0.794, "mrr_at_3": 0.8339999999999999, "mrr_at_5": 0.8417999999999999, "mrr_at_10": 0.8457571428571429, "mrr_at_20": 0.8470214830491145, "mrr_at_100": 0.847837729621968, "mrr_at_1000": 0.847903412604169, "naucs_at_1_max": 0.7301598401598397, "naucs_at_1_std": 0.13073426573426558, "naucs_at_1_diff1": 0.9151098901098901, "naucs_at_3_max": 0.7855154311167684, "naucs_at_3_std": 0.19653197581928136, "naucs_at_3_diff1": 0.8981466751511297, "naucs_at_5_max": 0.8119719432488443, "naucs_at_5_std": 0.1740338411787435, "naucs_at_5_diff1": 0.9070278517911231, "naucs_at_10_max": 0.8026658939437882, "naucs_at_10_std": 0.2013587044013021, "naucs_at_10_diff1": 0.9108149006729143, "naucs_at_20_max": 0.9162257495590788, "naucs_at_20_std": 0.37184873949579655, "naucs_at_20_diff1": 0.9110384894698595, "naucs_at_100_max": 0.947712418300658, "naucs_at_100_std": 0.3913165266106556, "naucs_at_100_diff1": 0.9738562091503188, "naucs_at_1000_max": 1.0, "naucs_at_1000_std": 1.0, "naucs_at_1000_diff1": 1.0}, "tabfquad_subsampled": {"ndcg_at_1": 0.82143, "ndcg_at_3": 0.88125, "ndcg_at_5": 0.88724, "ndcg_at_10": 0.89646, "ndcg_at_20": 0.90181, "ndcg_at_100": 0.90464, "ndcg_at_1000": 0.90464, "map_at_1": 0.82143, "map_at_3": 0.86726, "map_at_5": 0.87065, "map_at_10": 0.87444, "map_at_20": 0.87587, "map_at_100": 0.87633, "map_at_1000": 0.87633, "recall_at_1": 0.82143, "recall_at_3": 0.92143, "recall_at_5": 0.93571, "recall_at_10": 0.96429, "recall_at_20": 0.98571, "recall_at_100": 1.0, "recall_at_1000": 1.0, "precision_at_1": 0.82143, "precision_at_3": 0.30714, "precision_at_5": 0.18714, "precision_at_10": 0.09643, "precision_at_20": 0.04929, "precision_at_100": 0.01, "precision_at_1000": 0.001, "mrr_at_1": 0.8285714285714286, "mrr_at_3": 0.8708333333333336, "mrr_at_5": 0.8742261904761907, "mrr_at_10": 0.8779606009070298, "mrr_at_20": 0.8793680292629874, "mrr_at_100": 0.8798495301637429, "mrr_at_1000": 0.8798495301637429, "naucs_at_1_max": 0.4043094907866057, "naucs_at_1_std": 0.16813948880523036, "naucs_at_1_diff1": 0.8608282147810575, "naucs_at_3_max": 0.5415287327052029, "naucs_at_3_std": 0.32170443935149834, "naucs_at_3_diff1": 0.7899371869960117, "naucs_at_5_max": 0.5449216723726537, "naucs_at_5_std": 0.295959124390499, "naucs_at_5_diff1": 0.768025728810043, "naucs_at_10_max": 0.6852007469654555, "naucs_at_10_std": 0.596965452847805, "naucs_at_10_diff1": 0.7665732959850585, "naucs_at_20_max": 0.8231792717086873, "naucs_at_20_std": 0.7496498599439745, "naucs_at_20_diff1": 0.6626984126984181, "naucs_at_100_max": 1.0, "naucs_at_100_std": 1.0, "naucs_at_100_diff1": 1.0, "naucs_at_1000_max": NaN, "naucs_at_1000_std": NaN, "naucs_at_1000_diff1": NaN}, "tatdqa": {"ndcg_at_1": 0.62237, "ndcg_at_3": 0.72528, "ndcg_at_5": 0.7518, "ndcg_at_10": 0.7704, "ndcg_at_20": 0.77829, "ndcg_at_100": 0.78521, "ndcg_at_1000": 0.787, "map_at_1": 0.62237, "map_at_3": 0.70014, "map_at_5": 0.71487, "map_at_10": 0.72262, "map_at_20": 0.72486, "map_at_100": 0.72587, "map_at_1000": 0.72594, "recall_at_1": 0.62237, "recall_at_3": 0.79796, "recall_at_5": 0.8623, "recall_at_10": 0.91942, "recall_at_20": 0.95009, "recall_at_100": 0.98677, "recall_at_1000": 1.0, "precision_at_1": 0.62237, "precision_at_3": 0.26599, "precision_at_5": 0.17246, "precision_at_10": 0.09194, "precision_at_20": 0.0475, "precision_at_100": 0.00987, "precision_at_1000": 0.001, "mrr_at_1": 0.6277811184606134, "mrr_at_3": 0.7027460412908401, "mrr_at_5": 0.7170875927039494, "mrr_at_10": 0.7253199896916075, "mrr_at_20": 0.7277180395724787, "mrr_at_100": 0.7287256043001191, "mrr_at_1000": 0.7287991066528416, "naucs_at_1_max": 0.13553936519738788, "naucs_at_1_std": -0.3187519674953281, "naucs_at_1_diff1": 0.7778249427309417, "naucs_at_3_max": 0.2141299663362708, "naucs_at_3_std": -0.2863201435902689, "naucs_at_3_diff1": 0.6746349432069665, "naucs_at_5_max": 0.26199278239549034, "naucs_at_5_std": -0.2097766943253057, "naucs_at_5_diff1": 0.6504413592115734, "naucs_at_10_max": 0.33005867437923625, "naucs_at_10_std": 0.042015716590848955, "naucs_at_10_diff1": 0.5806124077911619, "naucs_at_20_max": 0.38884886089913584, "naucs_at_20_std": 0.25394752368146895, "naucs_at_20_diff1": 0.576841503912515, "naucs_at_100_max": 0.6555237795463111, "naucs_at_100_std": 0.7055850937111812, "naucs_at_100_diff1": 0.6635432021140835, "naucs_at_1000_max": NaN, "naucs_at_1000_std": NaN, "naucs_at_1000_diff1": NaN}, "shift_project": {"ndcg_at_1": 0.74, "ndcg_at_3": 0.84833, "ndcg_at_5": 0.85651, "ndcg_at_10": 0.86656, "ndcg_at_20": 0.86887, "ndcg_at_100": 0.87251, "ndcg_at_1000": 0.87251, "map_at_1": 0.74, "map_at_3": 0.82333, "map_at_5": 0.82783, "map_at_10": 0.83218, "map_at_20": 0.8327, "map_at_100": 0.83316, "map_at_1000": 0.83316, "recall_at_1": 0.74, "recall_at_3": 0.92, "recall_at_5": 0.94, "recall_at_10": 0.97, "recall_at_20": 0.98, "recall_at_100": 1.0, "recall_at_1000": 1.0, "precision_at_1": 0.74, "precision_at_3": 0.30667, "precision_at_5": 0.188, "precision_at_10": 0.097, "precision_at_20": 0.049, "precision_at_100": 0.01, "precision_at_1000": 0.001, "mrr_at_1": 0.76, "mrr_at_3": 0.8333333333333335, "mrr_at_5": 0.8378333333333334, "mrr_at_10": 0.8423571428571431, "mrr_at_20": 0.8428834586466167, "mrr_at_100": 0.84334179197995, "mrr_at_1000": 0.84334179197995, "naucs_at_1_max": -0.18346782368922665, "naucs_at_1_std": -0.312314991281781, "naucs_at_1_diff1": 0.6844410202343775, "naucs_at_3_max": 0.05304621848739506, "naucs_at_3_std": 0.02485994397759261, "naucs_at_3_diff1": 0.6601890756302518, "naucs_at_5_max": 0.30104263927793357, "naucs_at_5_std": 0.15561780267662587, "naucs_at_5_diff1": 0.7639277933395614, "naucs_at_10_max": -0.35434173669467856, "naucs_at_10_std": -0.3963585434173691, "naucs_at_10_diff1": 0.8638344226579531, "naucs_at_20_max": -0.5929038281979383, "naucs_at_20_std": -0.5088702147525547, "naucs_at_20_diff1": 0.9346405228758136, "naucs_at_100_max": NaN, "naucs_at_100_std": NaN, "naucs_at_100_diff1": NaN, "naucs_at_1000_max": NaN, "naucs_at_1000_std": NaN, "naucs_at_1000_diff1": NaN}}

special_tokens_map.json

@@ -0,0 +1,31 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer_config.json

@@ -0,0 +1,144 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "chat_template": "{% set image_count = namespace(value=0) %}{% set video_count = namespace(value=0) %}{% for message in messages %}{% if loop.first and message['role'] != 'system' %}<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n{% endif %}<|im_start|>{{ message['role'] }}\n{% if message['content'] is string %}{{ message['content'] }}<|im_end|>\n{% else %}{% for content in message['content'] %}{% if content['type'] == 'image' or 'image' in content or 'image_url' in content %}{% set image_count.value = image_count.value + 1 %}{% if add_vision_id %}Picture {{ image_count.value }}: {% endif %}<|vision_start|><|image_pad|><|vision_end|>{% elif content['type'] == 'video' or 'video' in content %}{% set video_count.value = video_count.value + 1 %}{% if add_vision_id %}Video {{ video_count.value }}: {% endif %}<|vision_start|><|video_pad|><|vision_end|>{% elif 'text' in content %}{{ content['text'] }}{% endif %}{% endfor %}<|im_end|>\n{% endif %}{% endfor %}{% if add_generation_prompt %}<|im_start|>assistant\n{% endif %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "model_max_length": 32768,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "left",
+  "processor_class": "ColQwen2Processor",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

training_config.yml

@@ -0,0 +1,43 @@
+config:
+  (): colpali_engine.trainer.colmodel_training.ColModelTrainingConfig
+  output_dir: !path ../../../models/colqwen2-multi
+  processor:
+    (): colpali_engine.utils.transformers_wrappers.AllPurposeWrapper
+    class_to_instanciate: !ext colpali_engine.models.ColQwen2Processor
+    pretrained_model_name_or_path:  "./models/Qwen2-VL-2B-Instruct" # "./models/paligemma-3b-mix-448"
+    # max_length: 50
+
+  model:
+    (): colpali_engine.utils.transformers_wrappers.AllPurposeWrapper
+    class_to_instanciate: !ext colpali_engine.models.ColQwen2
+    pretrained_model_name_or_path: "./models/colqwen2_base"
+    torch_dtype:  !ext torch.bfloat16
+    use_cache: false
+#    device_map: "auto"
+#    quantization_config:
+#      (): transformers.BitsAndBytesConfig
+#      load_in_4bit: true
+#      bnb_4bit_quant_type: "nf4"
+#      bnb_4bit_compute_dtype:  "bfloat16"
+#      bnb_4bit_use_double_quant: true
+
+  dataset_loading_func: !ext colpali_engine.utils.dataset_transformation.load_train_set_detailed
+  eval_dataset_loader: !import ../data/test_data.yaml
+
+  # max_length: 50
+  run_eval: true
+  add_suffix: true
+  loss_func:
+    (): colpali_engine.loss.late_interaction_losses.ColbertPairwiseCELoss
+  tr_args: !import ../tr_args/default_tr_args.yaml
+  peft_config:
+    (): peft.LoraConfig
+    r: 32
+    lora_alpha: 32
+    lora_dropout: 0.1
+    init_lora_weights: "gaussian"
+    bias: "none"
+    task_type: "FEATURE_EXTRACTION"
+    target_modules: '(.*(model).*(down_proj|gate_proj|up_proj|k_proj|q_proj|v_proj|o_proj).*$|.*(custom_text_proj).*$)'
+    # target_modules: '(.*(language_model).*(down_proj|gate_proj|up_proj|k_proj|q_proj|v_proj|o_proj).*$|.*(custom_text_proj).*$)'
+