Names a specimen crop at **subfamily, tribe, genus and species** for **400 bark and ambrosia beetle species** (115 genera, 26 tribes, 2 subfamilies), each level with a calibrated probability and a "known / unsure" decision, and reports the deepest level it is confident about: a species, "Genus sp. (species undetermined)", "Tribe (genus undetermined)", "Subfamily (tribe undetermined)" or "unrecognised". It is tuned to **avoid wrong names**: by default it stops at a coarser level rather than guess. The full species list is at the end of this card and in `species.csv`. Files: `model.safetensors` (weights), `deploy.safetensors` (kNN reference bank, validation score distributions), `config.json` (backbone, taxonomy, temperatures, thresholds, test report), `species.csv` (the 400 species), `calibration.md` (how the unknown check and thresholds were chosen). Classification of the benchmark's own species (top-1, no abstention): `iid_test` 650 scored specimens: subfamily 1.000, tribe 0.983, genus 0.922, species 0.838; all 5,017 `iid_test` specimens: genus 0.973, species 0.918; `inat_test` field photos: species 0.784.
This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated. Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations. Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).
We introduce Bamba-9B, a decoder-only language model based on the Mamba-2 architecture and is designed to handle a wide range of text generation tasks. It is trained from scratch using a two-stage training approach. In the first stage, the model is trained on 2 trillion tokens from the Dolma v1.7 dataset. In the second stage, it undergoes additional training on 200 billion tokens, leveraging a carefully curated blend of high-quality data to further refine its performance and enhance output quality. The current release includes the following models: | **Stage** | **Bamba 9B** | **Quantized** | **Note** | |----------------------|----------------------------------------------------------------------|-------------------------------------------------------------------------|-------------------------------------------------------------------| | **Base Model** | ibm-fms/Bamba-9B-v1 | ibm-fms/Bamba-9B-fp8 | Stage 2 pretraining | | **Base Model** | ibm-fms/Bamba-9B-2T | ibm-fms/Bamba-9B-fp8 | Stage 1 pretraining | | **Base Model** | ibm-fms/Bamba-9B-1.8T| ibm-fms/Bamba-9B-fp8 | Intermediate checkpoints during Stage 1, more to come | | **SFT** | coming soon | coming soon | to be released in the next drop | | **DPO** | coming soon | coming soon | to be released in the next drop | For users using our HF versions of the model, you would need to install the latest transformers which includes our newly merged implementation for our Bamba models:
**Model Summary:** Micro-G3.3-8B-Instruct-1B is a 1-billion parameter micro language model fine-tuned for reasoning and instruction-following capabilities. Built on top of Granite-3.3-8B-Instruct, with only 3 hidden layers, this model is trained to maximize performance and hardware compatibility at minimal compute cost. **Generation:** This is a simple example of how to use Micro-G3.3-8B-Instruct-1B model.
Llama3-MS-CLIP is the first Vision-Language Model in the CLIP family that understands multispectral imagery. It is trained on one million image-text pairs from the SSL4EO-S12-v1.1 dataset with generated captions. The CLIP model consists of two encoders for text and images. We extended the RGB patch embeddings to multispectral input and initialized the weights of the additional input channels with zeros. During the continual pre-training, the images and texts of each batch are encoded and combined. The loss increases the similarity of matching pairs while decreasing other combinations. We evaluated Llama3-MS-CLIP with zero-shot classification and text-to-image retrieval results, measured in accuracy (%) ↑ and mAP@100 (%) ↑, respectively. The following Figure compares our model with the OpenCLIP baselines and other EO-VLMs. We applied a smoothed min-max scaling and annotated the lowest and highest scores. Our multispectral CLIP model is outperforming other RGB-based models on most benchmarks.
TerraMind is the first multimodal any-to-any generative foundation model for Earth Observation jointly developed by IBM, ESA, and Forschungszentrum Jülich. TerraMind uses a dual-scale transformer-based encoder-decoder architecture, simultaneously processing pixel-level and token-level data. The model was pre-trained on 500B tokens from 9M spatiotemporally aligned multimodal samples from the TerraMesh dataset. Modality-specific patch embeddings allow direct processing of raw inputs, while modality-specific FSQ-VAEs are used for image tokenization. For sequence-like modalities such as coordinates, an adapted WordPiece tokenizer is employed. During pre-training, TerraMind leverages masked token reconstruction, learning complex cross-modal correlations to generate high-quality latent representations.
TerraMind is the first multimodal any-to-any generative foundation model for Earth Observation jointly developed by IBM, ESA, and Forschungszentrum Jülich. TerraMind uses a dual-scale transformer-based encoder-decoder architecture, simultaneously processing pixel-level and token-level data. The model was pre-trained on 500B tokens from 9M spatiotemporally aligned multimodal samples from the TerraMesh dataset. Modality-specific patch embeddings allow direct processing of raw inputs, while modality-specific FSQ-VAEs are used for image tokenization. For sequence-like modalities such as coordinates, an adapted WordPiece tokenizer is employed. During pre-training, TerraMind leverages masked token reconstruction, learning complex cross-modal correlations to generate high-quality latent representations.
TerraMind is the first multimodal any-to-any generative foundation model for Earth Observation jointly developed by IBM, ESA, and Forschungszentrum Jülich. TerraMind uses a dual-scale transformer-based encoder-decoder architecture, simultaneously processing pixel-level and token-level data. The model was pre-trained on 500B tokens from 9M spatiotemporally aligned multimodal samples from the TerraMesh dataset. Modality-specific patch embeddings allow direct processing of raw inputs, while modality-specific FSQ-VAEs are used for image tokenization. For sequence-like modalities such as coordinates, an adapted WordPiece tokenizer is employed. During pre-training, TerraMind leverages masked token reconstruction, learning complex cross-modal correlations to generate high-quality latent representations.
TerraMind is the first multimodal any-to-any generative foundation model for Earth Observation jointly developed by IBM, ESA, and Forschungszentrum Jülich. The model is pre-trained using FSQ-VAE tokens as targets. This tokenizer encodes and decodes land-use land-cover (LULC) maps for the TerraMind model. The tokenizer uses FSQ with five dimensions and a codebook size of 4'375 tokens. The model was pre-trained for 20 epochs on nine million LULC images from the TerraMesh dataset which are sourced from ESRI. The maps include nine classes and a 10th no-data class: No data, water, trees, flooded vegetation, crops, built area, bare ground, snow/ice, clouds, rangeland. The tokenizer is fully integrated into the fine-tuning toolkit TerraTorch. You can initialize the pre-trained tokenizer with:
TerraMind is the first multimodal any-to-any generative foundation model for Earth Observation jointly developed by IBM, ESA, and Forschungszentrum Jülich. The model is pre-trained using FSQ-VAE tokens as targets. This tokenizer encodes and decodes Sentinel-2 L2A satellite images for the TerraMind model. The tokenizer uses FSQ with five dimensions and a codebook size of 15'360 tokens. The decoding process uses diffusion steps for the reconstruction. The model was pre-trained for 100 epochs on nine million S-2 L2A images from the TerraMesh dataset. The tokenizer is fully integrated into the fine-tuning toolkit TerraTorch. You can initialize the pre-trained tokenizer with:
