wjbmattingly/lfm2-vl-450m-caroline

model_step_7000

Model Description

This model is a fine-tuned version of LiquidAI/LFM2-VL-450M using the brute-force-training package.

• Base Model: LiquidAI/LFM2-VL-450M
• Training Status: 🔄 In Progress
• Generated: 2025-08-18 20:39:32
• Training Steps: 7,000

Training Details

Dataset

• Dataset: CATMuS/medieval
• Training Examples: 148,000
• Validation Examples: 1,999

Training Configuration

• Max Steps: 10,000
• Batch Size: 2
• Learning Rate: 1e-05
• Gradient Accumulation: 1 steps
• Evaluation Frequency: Every 500 steps

Current Performance

• Training Loss: 0.619257
• Evaluation Loss: 0.722366

Pre-Training Evaluation

Initial Model Performance (before training):

• Loss: 1.297430
• Perplexity: 3.66
• Character Accuracy: 27.7%
• Word Accuracy: 11.6%

Evaluation History

All Checkpoint Evaluations

Step	Checkpoint Type	Loss	Perplexity	Char Acc	Word Acc	Improvement vs Pre
Pre	pre_training	1.2974	3.66	27.7%	11.6%	+0.0%
500	checkpoint	0.9454	2.57	39.4%	19.9%	+27.1%
1,000	checkpoint	0.8644	2.37	38.7%	19.1%	+33.4%
1,500	checkpoint	0.8402	2.32	38.4%	18.9%	+35.2%
2,000	checkpoint	0.8139	2.26	37.9%	19.8%	+37.3%
2,500	checkpoint	0.7890	2.20	38.5%	18.9%	+39.2%
3,000	checkpoint	0.7793	2.18	39.3%	19.5%	+39.9%
3,500	checkpoint	0.7639	2.15	42.7%	21.4%	+41.1%
4,000	checkpoint	0.7483	2.11	41.2%	20.4%	+42.3%
4,500	checkpoint	0.7466	2.11	37.3%	18.8%	+42.5%
5,000	checkpoint	0.7358	2.09	40.4%	20.5%	+43.3%
5,500	checkpoint	0.7321	2.08	38.1%	18.9%	+43.6%
6,000	checkpoint	0.7276	2.07	38.8%	17.6%	+43.9%
6,500	checkpoint	0.7190	2.05	41.5%	18.9%	+44.6%
7,000	checkpoint	0.7224	2.06	41.6%	18.7%	+44.3%

Training Progress

Recent Training Steps (Loss Only)

Step	Training Loss	Timestamp
6,991	0.846698	2025-08-18T20:39
6,992	0.538150	2025-08-18T20:39
6,993	0.721188	2025-08-18T20:39
6,994	0.819544	2025-08-18T20:39
6,995	0.925656	2025-08-18T20:39
6,996	0.724563	2025-08-18T20:39
6,997	0.738329	2025-08-18T20:39
6,998	0.658910	2025-08-18T20:39
6,999	0.439738	2025-08-18T20:39
7,000	0.619257	2025-08-18T20:39

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_7000")
tokenizer = AutoTokenizer.from_pretrained("./model_step_7000")

# Your inference code here

Training Configuration

{
  "dataset_name": "CATMuS/medieval",
  "model_name": "LiquidAI/LFM2-VL-450M",
  "max_steps": 10000,
  "eval_steps": 500,
  "num_accumulation_steps": 1,
  "learning_rate": 1e-05,
  "train_batch_size": 2,
  "val_batch_size": 2,
  "train_select_start": 0,
  "train_select_end": 148000,
  "val_select_start": 148001,
  "val_select_end": 150000,
  "train_field": "train",
  "val_field": "train",
  "image_column": "im",
  "text_column": "text",
  "user_text": "Transcribe this medieval manuscript line.",
  "max_image_size": 200
}

Model Card Metadata

• Base Model: LiquidAI/LFM2-VL-450M
• 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 2025-08-18 20:39:32