Upload folder using huggingface_hub

.gitattributes

@@ -0,0 +1,3 @@
+adapter_model.safetensors filter=lfs diff=lfs merge=lfs -text
+optimizer.pt filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

README.md

@@ -8,151 +8,202 @@ tags:
 - transformers
 - trl
 - unsloth
-license: cc-by-4.0
-datasets:
-- NovaIALATAM/CuPer_Text
-- NovaIALATAM/CuPer_Images
-language:
-- es
-pipeline_tag: image-text-to-text
 ---
 
-![AMARU](./amaru-vl.jpg)
-
-# BIENVENIDOS A AMARU
-
-## Detalles del Modelo
-- **Nombre del Modelo:** Amaru-VL-3B
-- **Versión del Modelo:** 1.0
-- **Desarrollador/Organización:** NovaIA, laboratorio de IA de Grupo Neura
-- **Fecha de Lanzamiento:** Agosto 2025
-- **Licencia:** cc-by-4.0
-- **Idioma(s):** Español (principalmente, con capacidades multilingües heredadas del modelo base)
-- **Repositorio:** [NovaIALATAM/Amaru-VL-3B](https://huggingface.co/NovaIALATAM/Amaru-VL-3B)
-- **Contacto:** Para más información, contactar a NovaIA a través de [su sitio web](https://www.gruponeura.ia).
-
-## Descripción del Modelo
-Amaru-VL-3B es un modelo de lenguaje multimodal (VLM) fine-tuneado mediante LoRA a partir del modelo base Qwen2.5-VL-3B. Desarrollado por NovaIA, el laboratorio de inteligencia artificial de Grupo Neura, este modelo está diseñado para ser un experto en las culturas precolombinas del Perú, abordando las limitaciones de los grandes modelos de lenguaje que a menudo cometen errores en tareas relacionadas con el patrimonio cultural latinoamericano. Amaru representa el primer acercamiento hacia un LLM latinoamericano especializado, manteniendo las capacidades generales de generación de texto del modelo base para tareas de propósito general, mientras integra conocimiento profundo sobre civilizaciones ancestrales peruanas como Chavín, Moche, Nazca, Chimú, Lambayeque y otras.
-
-El modelo soporta entradas multimodales (texto e imágenes) y es útil para tareas como análisis de artefactos arqueológicos, respuestas a preguntas históricas, generación de descripciones culturales y más. Fue entrenado con un total de aproximadamente 5 millones de tokens, combinando datasets abiertos y privados para asegurar precisión histórica y cultural.
-
-**Arquitectura:** Basado en Qwen2.5-VL-3B, un modelo de visión-lenguaje con aproximadamente 3 mil millones de parámetros, optimizado para eficiencia con soporte para cuantización (e.g., 4-bit) y long context.
-
-**Objetivos Principales:**
-- Proporcionar conocimiento preciso sobre culturas precolombinas peruanas.
-- Servir como base para aplicaciones educativas, de investigación y de preservación cultural en América Latina.
-- Mantener versatilidad para generación de texto general.
-
-## Usos
-### Usos Directos
-- **Análisis Multimodal:** Identificación y descripción de imágenes relacionadas con artefactos, arquitectura o iconografía precolombina (e.g., máscaras Lambayeque, cerámica Moche).
-- **Preguntas y Respuestas:** Consultas sobre historia, rituales, economía y arte de civilizaciones peruanas.
-- **Generación de Texto:** Creación de contenido educativo o narrativo sobre patrimonio cultural, así como tareas generales de lenguaje.
-
-### Ejemplo de Carga y Uso
-El modelo se carga fácilmente con Unsloth para inferencia eficiente. Aquí un ejemplo:
-
-```python
-from unsloth import FastVisionModel, FastTextModel
-import torch
-from transformers import TextStreamer
-from PIL import Image
-import requests
-from io import BytesIO
-
-model_id = "NovaIALATAM/Amaru-VL-3B"
-
-# Cargar modelo y tokenizer con Unsloth
-base_model, tokenizer = FastVisionModel.from_pretrained(
-    model_name=model_id,
-    max_seq_length=8192,
-    dtype=None,
-    load_in_4bit=False,  
-    use_gradient_checkpointing="unsloth", 
-)
-
-FastVisionModel.for_inference(base_model)
-
-# Imagen de ejemplo
-path_image = "https://cloudfront-us-east-1.images.arcpublishing.com/infobae/WARBZQ3A65BQRLN3D63I3JJBIY.jpg"
-response = requests.get(path_image, timeout=10)
-response.raise_for_status()
-image = Image.open(BytesIO(response.content))
-
-messages = [
-    {"role": "system", "content": [
-        {"type": "text", "text": "Amaru, eres un experto conocedor de diferentes culturas precolombinas del Perú."}]},
-    {"role": "user", "content": [
-        {"type": "image", "image": image},
-        {"type": "text", "text": "¿Qué es esto?"}
-    ]}
-]
-
-inputs = tokenizer(
-    images=image,
-    text=tokenizer.apply_chat_template(messages, add_generation_prompt=True),
-    add_special_tokens=False,
-    return_tensors="pt",
-).to("cuda")
-
-text_streamer = TextStreamer(tokenizer, skip_prompt=True)
-_ = base_model.generate(**inputs, streamer=text_streamer, max_new_tokens=2048,
-                   use_cache=True, temperature=0.05, min_p=0.1)
-```
-
-### Usos en Aplicaciones Downstream
-- Integración en chatbots educativos para museos o plataformas de aprendizaje.
-- Fine-tuning adicional para tareas específicas en patrimonio cultural o turismo.
-- Uso en herramientas de IA para investigación arqueológica.
-
-### Usos Fuera de Alcance
-- No se recomienda para aplicaciones médicas, legales o de alta precisión sin validación adicional.
-- Evitar usos que promuevan desinformación histórica o cultural.
-- No está optimizado para idiomas no latinos o contextos no relacionados con el conocimiento base.
-
-## Sesgos, Riesgos y Limitaciones
-- **Sesgos:** El modelo hereda sesgos potenciales del modelo base Qwen2.5-VL-3B y de los datasets de entrenamiento. Dado el enfoque en culturas peruanas, podría subrepresentar otras culturas latinoamericanas o globales. Los datasets abiertos (CuPer_Text y CuPer_Images) fueron curados para precisión, pero los privados podrían introducir sesgos no documentados.
-- **Riesgos:** Posibles alucinaciones en respuestas fuera del dominio precolombino, como en modelos de IA generales. En temas sensibles como rituales indígenas, podría generar contenido inexacto si no se guía adecuadamente.
-- **Limitaciones:** 
-  - Rendimiento inferior en idiomas no españoles o en contextos modernos no relacionados.
-  - Requiere hardware GPU para inferencia eficiente (soporte para cuantización mitiga esto).
-  - No evaluado exhaustivamente; se recomienda validación humana para usos críticos.
-- **Recomendaciones:**
-  - Usar prompts específicos (e.g., "Actúa como un experto en culturas precolombinas del Perú") para maximizar precisión.
-  - Usar una temperatura baja para evitar alucinaciones: temperature = 0.05
-
-## Detalles del Entrenamiento
-### Datos de Entrenamiento
-El modelo fue fine-tuneado con un total de ~5 millones de tokens, incluyendo:
-- **CuPer_Text** (abierto, cc-by-nc-4.0): 4,144 muestras (3,268 train, 876 test), ~650,000 tokens. Formato: Preguntas y respuestas en texto sobre culturas precolombinas peruanas (e.g., historia, arquitectura, rituales). Creado por NovaIA para asegurar precisión cultural.
-- **CuPer_Images** (abierto, cc-by-nc-4.0): 4,247 muestras (3,396 train, 851 test), ~2.2 millones de tokens. Formato: Preguntas, imágenes y respuestas en texto, enfocadas en aspectos visuales como artefactos, estructuras y iconografía (e.g., acueductos Nazca, máscaras Lambayeque).
-- **Datasets Privados:** Datos adicionales curados por NovaIA, contribuyendo al total de 5 millones de tokens, con énfasis en conocimiento histórico validado.
-
-Todos los datos están en español y centrados en culturas como Lambayeque, Moche, Nazca, Chavín y Chimú.
-
-### Procedimiento de Entrenamiento
-- **Método:** Fine-tuning LoRA (Low-Rank Adaptation) sobre Qwen2.5-VL-3B para eficiencia.
-- **Hiperparámetros:** No especificados públicamente; optimizados para contexto largo (hasta 8192 tokens) y multimodalidad.
-- **Hardware:** Entrenado en infraestructura de NovaIA, compatible con Unsloth para aceleración.
-- **Duración:** No detallada, pero enfocado en precisión cultural sobre volumen.
-
-## Impacto Ambiental
-- **Estimación de Carbono:** El fine-tuning con LoRA es eficiente, estimando un impacto bajo (~equivalente a unas horas de GPU A100). No se calculó formalmente.
-- **Eficiencia:** Soporte para cuantización 4-bit reduce el consumo en inferencia.
-
-## Citación
-Si utilizas este modelo en tu investigación o aplicación, por favor cita como:
-```
-@misc{amaru_vl_3b_2025,
-  title = {Amaru-VL-3B: Modelo Multimodal Especializado en Culturas Precolombinas del Perú},
-  author = {NovaIA - Grupo Neura},
-  year = {2025},
-  publisher = {Hugging Face},
-  url = {https://huggingface.co/NovaIALATAM/Amaru-VL-3B}
-}
-```
-
-## Más Información
-Para más detalles sobre los datasets:
-- [CuPer_Text](https://huggingface.co/datasets/NovaIALATAM/CuPer_Text)
-- [CuPer_Images](https://huggingface.co/datasets/NovaIALATAM/CuPer_Images)
+# 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 0.17.0

adapter_config.json

@@ -0,0 +1,45 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": {
+    "base_model_class": "Qwen2_5_VLForConditionalGeneration",
+    "parent_library": "transformers.models.qwen2_5_vl.modeling_qwen2_5_vl"
+  },
+  "base_model_name_or_path": "unsloth/Qwen2.5-VL-3B-Instruct",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 4,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "qalora_group_size": 16,
+  "r": 2,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "k_proj",
+    "up_proj",
+    "gate_proj",
+    "v_proj",
+    "o_proj",
+    "down_proj",
+    "q_proj"
+  ],
+  "target_parameters": null,
+  "task_type": null,
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:385ceb02b804f3c2831c7943305ed5f27c2134e543b1464903990707eddd0371
+size 18676192

added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

chat_template.jinja

@@ -0,0 +1,7 @@
+{% 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
+You are a helpful assistant.<|im_end|>
+{% endif %}<|im_start|>{{ message['role'] }}
+{% if message['content'] is string %}{{ message['content'] }}<|im_end|>
+{% 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|>
+{% endif %}{% endfor %}{% if add_generation_prompt %}<|im_start|>assistant
+{% endif %}

optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac2cb766e5373a5c6c5bbb4a7008f0942d3f0122b65cfb1c62a3ff451de28a7a
+size 8363492

preprocessor_config.json

@@ -0,0 +1,37 @@
+{
+  "crop_size": null,
+  "data_format": "channels_first",
+  "default_to_square": true,
+  "device": null,
+  "disable_grouping": null,
+  "do_center_crop": null,
+  "do_convert_rgb": true,
+  "do_normalize": true,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.48145466,
+    0.4578275,
+    0.40821073
+  ],
+  "image_processor_type": "Qwen2VLImageProcessorFast",
+  "image_std": [
+    0.26862954,
+    0.26130258,
+    0.27577711
+  ],
+  "input_data_format": null,
+  "max_pixels": 12845056,
+  "merge_size": 2,
+  "min_pixels": 3136,
+  "patch_size": 14,
+  "processor_class": "Qwen2_5_VLProcessor",
+  "resample": 3,
+  "rescale_factor": 0.00392156862745098,
+  "return_tensors": null,
+  "size": {
+    "longest_edge": 12845056,
+    "shortest_edge": 3136
+  },
+  "temporal_patch_size": 2
+}

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": "<|vision_pad|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1296e28e96c0a57e3bbee1bcdc014e39b2dde71f015c7e19e56cc4399de40c66
+size 11422166

tokenizer_config.json

@@ -0,0 +1,209 @@
+{
+  "add_bos_token": false,
+  "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
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "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,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 128000,
+  "pad_token": "<|vision_pad|>",
+  "padding_side": "right",
+  "processor_class": "Qwen2_5_VLProcessor",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

trainer_state.json

@@ -0,0 +1,334 @@
+{
+  "best_global_step": 200,
+  "best_metric": 0.8998275399208069,
+  "best_model_checkpoint": "/content/drive/MyDrive/amaru-txt-epoch-6/checkpoint-200",
+  "epoch": 0.966183574879227,
+  "eval_steps": 10,
+  "global_step": 200,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.04830917874396135,
+      "grad_norm": 1.9938534498214722,
+      "learning_rate": 3.5389440000000007e-05,
+      "loss": 0.8315,
+      "step": 10
+    },
+    {
+      "epoch": 0.04830917874396135,
+      "eval_loss": 0.9170326590538025,
+      "eval_runtime": 89.5861,
+      "eval_samples_per_second": 9.778,
+      "eval_steps_per_second": 0.156,
+      "step": 10
+    },
+    {
+      "epoch": 0.0966183574879227,
+      "grad_norm": 2.8435614109039307,
+      "learning_rate": 7.471104000000001e-05,
+      "loss": 0.7156,
+      "step": 20
+    },
+    {
+      "epoch": 0.0966183574879227,
+      "eval_loss": 0.92298823595047,
+      "eval_runtime": 89.6583,
+      "eval_samples_per_second": 9.77,
+      "eval_steps_per_second": 0.156,
+      "step": 20
+    },
+    {
+      "epoch": 0.14492753623188406,
+      "grad_norm": 2.7325339317321777,
+      "learning_rate": 7.819458354458255e-05,
+      "loss": 0.8015,
+      "step": 30
+    },
+    {
+      "epoch": 0.14492753623188406,
+      "eval_loss": 0.9262471199035645,
+      "eval_runtime": 89.6974,
+      "eval_samples_per_second": 9.766,
+      "eval_steps_per_second": 0.156,
+      "step": 30
+    },
+    {
+      "epoch": 0.1932367149758454,
+      "grad_norm": 2.380279779434204,
+      "learning_rate": 7.665694808929027e-05,
+      "loss": 0.7263,
+      "step": 40
+    },
+    {
+      "epoch": 0.1932367149758454,
+      "eval_loss": 0.9168504476547241,
+      "eval_runtime": 89.6787,
+      "eval_samples_per_second": 9.768,
+      "eval_steps_per_second": 0.156,
+      "step": 40
+    },
+    {
+      "epoch": 0.24154589371980675,
+      "grad_norm": 2.8440909385681152,
+      "learning_rate": 7.406804077083218e-05,
+      "loss": 0.7098,
+      "step": 50
+    },
+    {
+      "epoch": 0.24154589371980675,
+      "eval_loss": 0.9175106883049011,
+      "eval_runtime": 89.6725,
+      "eval_samples_per_second": 9.769,
+      "eval_steps_per_second": 0.156,
+      "step": 50
+    },
+    {
+      "epoch": 0.2898550724637681,
+      "grad_norm": 2.432560920715332,
+      "learning_rate": 7.050075887179768e-05,
+      "loss": 0.7212,
+      "step": 60
+    },
+    {
+      "epoch": 0.2898550724637681,
+      "eval_loss": 0.9199467301368713,
+      "eval_runtime": 89.7174,
+      "eval_samples_per_second": 9.764,
+      "eval_steps_per_second": 0.156,
+      "step": 60
+    },
+    {
+      "epoch": 0.33816425120772947,
+      "grad_norm": 3.1470143795013428,
+      "learning_rate": 6.605554830418061e-05,
+      "loss": 0.7116,
+      "step": 70
+    },
+    {
+      "epoch": 0.33816425120772947,
+      "eval_loss": 0.9214985966682434,
+      "eval_runtime": 89.6568,
+      "eval_samples_per_second": 9.771,
+      "eval_steps_per_second": 0.156,
+      "step": 70
+    },
+    {
+      "epoch": 0.3864734299516908,
+      "grad_norm": 3.123927354812622,
+      "learning_rate": 6.085757529877134e-05,
+      "loss": 0.7162,
+      "step": 80
+    },
+    {
+      "epoch": 0.3864734299516908,
+      "eval_loss": 0.9215625524520874,
+      "eval_runtime": 89.6808,
+      "eval_samples_per_second": 9.768,
+      "eval_steps_per_second": 0.156,
+      "step": 80
+    },
+    {
+      "epoch": 0.43478260869565216,
+      "grad_norm": 2.565199851989746,
+      "learning_rate": 5.5053202030981025e-05,
+      "loss": 0.806,
+      "step": 90
+    },
+    {
+      "epoch": 0.43478260869565216,
+      "eval_loss": 0.9178029298782349,
+      "eval_runtime": 89.7041,
+      "eval_samples_per_second": 9.765,
+      "eval_steps_per_second": 0.156,
+      "step": 90
+    },
+    {
+      "epoch": 0.4830917874396135,
+      "grad_norm": 2.8664753437042236,
+      "learning_rate": 4.880586542083376e-05,
+      "loss": 0.7107,
+      "step": 100
+    },
+    {
+      "epoch": 0.4830917874396135,
+      "eval_loss": 0.9137737154960632,
+      "eval_runtime": 89.6956,
+      "eval_samples_per_second": 9.766,
+      "eval_steps_per_second": 0.156,
+      "step": 100
+    },
+    {
+      "epoch": 0.5314009661835749,
+      "grad_norm": 2.94624400138855,
+      "learning_rate": 4.229147515001422e-05,
+      "loss": 0.7115,
+      "step": 110
+    },
+    {
+      "epoch": 0.5314009661835749,
+      "eval_loss": 0.913902223110199,
+      "eval_runtime": 89.6611,
+      "eval_samples_per_second": 9.77,
+      "eval_steps_per_second": 0.156,
+      "step": 110
+    },
+    {
+      "epoch": 0.5797101449275363,
+      "grad_norm": 3.233154296875,
+      "learning_rate": 3.569346047652783e-05,
+      "loss": 0.7752,
+      "step": 120
+    },
+    {
+      "epoch": 0.5797101449275363,
+      "eval_loss": 0.9105567336082458,
+      "eval_runtime": 89.6726,
+      "eval_samples_per_second": 9.769,
+      "eval_steps_per_second": 0.156,
+      "step": 120
+    },
+    {
+      "epoch": 0.6280193236714976,
+      "grad_norm": 2.791991949081421,
+      "learning_rate": 2.9197605316528352e-05,
+      "loss": 0.8082,
+      "step": 130
+    },
+    {
+      "epoch": 0.6280193236714976,
+      "eval_loss": 0.9074307084083557,
+      "eval_runtime": 89.6733,
+      "eval_samples_per_second": 9.769,
+      "eval_steps_per_second": 0.156,
+      "step": 130
+    },
+    {
+      "epoch": 0.6763285024154589,
+      "grad_norm": 2.9703595638275146,
+      "learning_rate": 2.2986817024745032e-05,
+      "loss": 0.8185,
+      "step": 140
+    },
+    {
+      "epoch": 0.6763285024154589,
+      "eval_loss": 0.9059156775474548,
+      "eval_runtime": 89.6839,
+      "eval_samples_per_second": 9.768,
+      "eval_steps_per_second": 0.156,
+      "step": 140
+    },
+    {
+      "epoch": 0.7246376811594203,
+      "grad_norm": 2.797487735748291,
+      "learning_rate": 1.7235976171826803e-05,
+      "loss": 0.8013,
+      "step": 150
+    },
+    {
+      "epoch": 0.7246376811594203,
+      "eval_loss": 0.9037850499153137,
+      "eval_runtime": 89.6764,
+      "eval_samples_per_second": 9.768,
+      "eval_steps_per_second": 0.156,
+      "step": 150
+    },
+    {
+      "epoch": 0.7729468599033816,
+      "grad_norm": 2.6389381885528564,
+      "learning_rate": 1.210701233624601e-05,
+      "loss": 0.8531,
+      "step": 160
+    },
+    {
+      "epoch": 0.7729468599033816,
+      "eval_loss": 0.9019114971160889,
+      "eval_runtime": 89.6735,
+      "eval_samples_per_second": 9.769,
+      "eval_steps_per_second": 0.156,
+      "step": 160
+    },
+    {
+      "epoch": 0.821256038647343,
+      "grad_norm": 2.7340869903564453,
+      "learning_rate": 7.744344564388566e-06,
+      "loss": 0.8288,
+      "step": 170
+    },
+    {
+      "epoch": 0.821256038647343,
+      "eval_loss": 0.9006927609443665,
+      "eval_runtime": 89.6331,
+      "eval_samples_per_second": 9.773,
+      "eval_steps_per_second": 0.156,
+      "step": 170
+    },
+    {
+      "epoch": 0.8695652173913043,
+      "grad_norm": 2.7714426517486572,
+      "learning_rate": 4.270814884295176e-06,
+      "loss": 0.8686,
+      "step": 180
+    },
+    {
+      "epoch": 0.8695652173913043,
+      "eval_loss": 0.8999435901641846,
+      "eval_runtime": 89.7018,
+      "eval_samples_per_second": 9.766,
+      "eval_steps_per_second": 0.156,
+      "step": 180
+    },
+    {
+      "epoch": 0.9178743961352657,
+      "grad_norm": 3.1497857570648193,
+      "learning_rate": 1.7842293753365276e-06,
+      "loss": 0.8859,
+      "step": 190
+    },
+    {
+      "epoch": 0.9178743961352657,
+      "eval_loss": 0.9000065326690674,
+      "eval_runtime": 89.6862,
+      "eval_samples_per_second": 9.767,
+      "eval_steps_per_second": 0.156,
+      "step": 190
+    },
+    {
+      "epoch": 0.966183574879227,
+      "grad_norm": 2.9853219985961914,
+      "learning_rate": 3.546041888197535e-07,
+      "loss": 0.8611,
+      "step": 200
+    },
+    {
+      "epoch": 0.966183574879227,
+      "eval_loss": 0.8998275399208069,
+      "eval_runtime": 89.6735,
+      "eval_samples_per_second": 9.769,
+      "eval_steps_per_second": 0.156,
+      "step": 200
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 207,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.99869509435392e+16,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

video_preprocessor_config.json

@@ -0,0 +1,43 @@
+{
+  "crop_size": null,
+  "data_format": "channels_first",
+  "default_to_square": true,
+  "device": null,
+  "do_center_crop": null,
+  "do_convert_rgb": true,
+  "do_normalize": true,
+  "do_pad": null,
+  "do_rescale": true,
+  "do_resize": true,
+  "do_sample_frames": false,
+  "fps": null,
+  "image_mean": [
+    0.48145466,
+    0.4578275,
+    0.40821073
+  ],
+  "image_std": [
+    0.26862954,
+    0.26130258,
+    0.27577711
+  ],
+  "input_data_format": null,
+  "max_frames": 768,
+  "max_pixels": 12845056,
+  "merge_size": 2,
+  "min_frames": 4,
+  "min_pixels": 3136,
+  "num_frames": null,
+  "patch_size": 14,
+  "processor_class": "Qwen2_5_VLProcessor",
+  "resample": 3,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "longest_edge": 12845056,
+    "shortest_edge": 3136
+  },
+  "size_divisor": null,
+  "temporal_patch_size": 2,
+  "video_metadata": null,
+  "video_processor_type": "Qwen2VLVideoProcessor"
+}