wjbmattingly/lfm2-vl-1.6B-catmus

model_step_10000

Model Description

This model is a fine-tuned version of LiquidAI/LFM2-VL-1.6B using the brute-force-training package.

• Base Model: LiquidAI/LFM2-VL-1.6B
• Training Status: 🔄 In Progress
• Generated: 2025-08-13 13:56:59
• Training Steps: 10,000

Training Details

Dataset

• Dataset: CATMuS/medieval
• Training Examples: 120,000
• Validation Examples: 29,999

Training Configuration

• Max Steps: 100,000
• Batch Size: 10
• Learning Rate: 1e-05
• Gradient Accumulation: 4 steps
• Evaluation Frequency: Every 10,000 steps

Current Performance

• Training Loss: 0.477249
• Evaluation Loss: 0.725169

Pre-Training Evaluation

Initial Model Performance (before training):

• Loss: 6.212058
• Perplexity: 498.73
• Character Accuracy: 18.6%
• Word Accuracy: 2.0%

Evaluation History

All Checkpoint Evaluations

Step	Checkpoint Type	Loss	Perplexity	Char Acc	Word Acc	Improvement vs Pre
Pre	pre_training	6.2121	498.73	18.6%	2.0%	+0.0%
10,000	checkpoint	0.7252	2.07	18.1%	1.4%	+88.3%

Training Progress

Recent Training Steps (Loss Only)

Step	Training Loss	Timestamp
9,991	0.889550	2025-08-13T13:54
9,992	0.641465	2025-08-13T13:54
9,993	0.997256	2025-08-13T13:54
9,994	0.746186	2025-08-13T13:54
9,995	0.850397	2025-08-13T13:54
9,996	0.359374	2025-08-13T13:54
9,997	1.091660	2025-08-13T13:54
9,998	1.327502	2025-08-13T13:54
9,999	0.802447	2025-08-13T13:54
10,000	0.477249	2025-08-13T13:54

Training Visualizations

Training Progress and Evaluation Metrics

This chart shows the training loss progression, character accuracy, word accuracy, and perplexity over time. Red dots indicate evaluation checkpoints.

Evaluation Comparison Across All Checkpoints

Comprehensive comparison of all evaluation metrics across training checkpoints. Red=Pre-training, Blue=Checkpoints, Green=Final.

Available Visualization Files:

• training_curves.png - 4-panel view: Training loss with eval points, Character accuracy, Word accuracy, Perplexity
• evaluation_comparison.png - 4-panel comparison: Loss, Character accuracy, Word accuracy, Perplexity across all checkpoints

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
# For vision-language models, use appropriate imports

model = AutoModelForCausalLM.from_pretrained("./model_step_10000")
tokenizer = AutoTokenizer.from_pretrained("./model_step_10000")

# Your inference code here

Training Configuration

{
  "dataset_name": "CATMuS/medieval",
  "model_name": "LiquidAI/LFM2-VL-1.6B",
  "max_steps": 100000,
  "eval_steps": 10000,
  "num_accumulation_steps": 4,
  "learning_rate": 1e-05,
  "train_batch_size": 10,
  "val_batch_size": 10,
  "train_select_start": 0,
  "train_select_end": 120000,
  "val_select_start": 120001,
  "val_select_end": 150000,
  "train_field": "train",
  "val_field": "train",
  "image_column": "im",
  "text_column": "text",
  "user_text": "Transcribe this medieval manuscript line"
}

Model Card Metadata

• Base Model: LiquidAI/LFM2-VL-1.6B
• Training Framework: brute-force-training
• Training Type: Fine-tuning
• License: Inherited from base model
• Language: Inherited from base model

---

This model card was automatically generated by brute-force-training on 2025-08-13 13:56:59