Update README.md

README.md

@@ -22,7 +22,7 @@ tags:
 <img src="https://www.grag.ai/wp-content/uploads/2024/12/GRAG-ICON-TO-WORDLOGO-Animation_Loop-small-ezgif.com-video-to-gif-converter.gif" alt="GRAG Logo" width="400" style="margin-left:'auto' margin-right:'auto' display:'block'"/>
 
 
-# Model Card for GRAG-LLAMA-3.1-8B-SFT-HESSIAN-AI
+# Model Card for GRAG-NEMO-12B-SFT-HESSIAN-AI
 
 <!-- Provide a quick summary of what the model is/does. -->
 
@@ -35,9 +35,9 @@ Our GRAG-LLAMA-SFT model are trained on this **[GRAG-SFT](https://huggingface.co
 The core models released in this batch are the following: 
 | Size | Training Tokens | 
 |------|--------|
-| [GRAG-LLAMA-CPT](https://huggingface.co/avemio/GRAG-LLAMA-3.1-8B-CPT-HESSIAN-AI)   | 507.47 million |
-| [GRAG-LLAMA-SFT](https://huggingface.co/avemio/GRAG-LLAMA-3.1-8B-SFT-HESSIAN-AI) |  2.03 billion  |  
-| [GRAG-LLAMA-ORPO](https://huggingface.co/avemio/GRAG-LLAMA-3.1-8B-ORPO-HESSIAN-AI) |  2.0577 billion  | 
+| [GRAG-NEMO-CPT](https://huggingface.co/avemio/GRAG-NEMO-12B-CPT-HESSIAN-AI)   | 507.47 million |
+| [GRAG-NEMO-SFT](https://huggingface.co/avemio/GRAG-NEMO-12B-SFT-HESSIAN-AI) |  2.03 billion  |  
+| [GRAG-NEMO-ORPO](https://huggingface.co/avemio/GRAG-NEMO-12B-ORPO-HESSIAN-AI) |  2.0577 billion  | 
 ### Model Description
 
 <!-- Provide a longer summary of what this model is. -->
@@ -71,7 +71,7 @@ Now, proceed as usual with HuggingFace:
 ```python
 from transformers import AutoModelForCausalLM, AutoTokenizer
  
-model_name = "avemio/GRAG-LLAMA-3.1-8B-SFT-HESSIAN-AI"
+model_name = "avemio/GRAG-NEMO-12B-SFT-HESSIAN-AI"
  
 model = AutoModelForCausalLM.from_pretrained(
     model_name,
@@ -133,7 +133,7 @@ Four evaluation metrics were employed across all subsets: language quality, over
 -   **Overall score:** This metric combined the results from the previous three metrics, offering a comprehensive evaluation of the model's capabilities across all subsets.
 
 
-| Metric                                    | [Vanila-Phi-3.5-Mini-4B](https://huggingface.co/microsoft/Phi-3.5-mini-instruct) | [GRAG-PHI-SFT](https://huggingface.co/avemio/GRAG-PHI-3.5-MINI-4B-SFT-HESSIAN-AI) | [GRAG-PHI-ORPO](https://huggingface.co/avemio/GRAG-PHI-3.5-MINI-4B-ORPO-HESSIAN-AI) | [GRAG-PHI-MERGED]() | GPT-3.5-TURBO | 
+| Metric                                    | [Vanila-Mistral-Nemo-Instruct-2407](https://huggingface.co/mistralai/Mistral-Nemo-Instruct-2407) | [GRAG-NEMO-SFT](https://huggingface.co/avemio/GRAG-NEMO-12B-SFT-HESSIAN-AI) | [GRAG-NEMO-ORPO](https://huggingface.co/avemio/GRAG-NEMO-12B-ORPO-HESSIAN-AI) | [GRAG-NEMO-MERGED]() | GPT-3.5-TURBO | 
 |------------------------------------------|---------------------------------------------------------------------------------|--------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------|-----------------------------|----------------|
 | **Average_language_quality**             | 85.88                                                                          | 89.61                                                                          | 89.1                                                                                          |                             |                |
 | **extraction_recall_weighted_overall_score**       | 35.2                                                                           | 52.3                                                                           | 48.8                                                                                          |                             |                |
@@ -162,25 +162,25 @@ The implementation of these tasks within RAG systems can significantly improve o
 ### Architecture
 
 
-| Parameter            | GRAG-LLAMA-SFT                                                                                   |
+| Parameter            | GRAG-NEMO-SFT                                                                                   |
 |-----------------------|-----------------------------------------------------------------------------------------------|
-| **d_model**          | 3072                                                                                          |
+| **d_model**          | 5120                                                                                         |
 | **num heads**        | 32                                                                                            |
-| **num layers**       | 32                                                                                            |
-| **MLP ratio**        | 3.5                                                                                          |
+| **num layers**       | 40                                                                                            |
+| **MLP ratio**        | 2.8                                                                                         |
 | **LayerNorm type**   | RMSNorm                                                                                       |
 | **pos embeddings**   | RoPE                                                                                          |
 | **attention variant**| Standard Multi-Head Self Attention                              |
 | **biases**           | none                                                                                          |
 | **block type**       | sequential                                                                                    |
 | **activation**       | SiLU                                                                                          |
-| **sequence length**  | 131072                                                                                        |
+| **sequence length**  | 1024000                                                                                      |
 | **weight typing**     | bfloat16  
 
 ### Hyperparameters 
 
 
-| Parameter                 | GRAG-LLAMA-SFT       |
+| Parameter                 | GRAG-NEMO-SFT       |
 |---------------------------|--------------------|
 | **warmup steps**          | 50                |
 | **peak LR**               | 5.0E-07           |
@@ -191,19 +191,19 @@ The implementation of these tasks within RAG systems can significantly improve o
 
 ## Environmental Impact
 
-GRAG-PHI-SFT, running on NVIDIA A100 with 8 GPUs for 5 days, has an approximate power consumption as follows:
+GRAG-NEMO-SFT, running on NVIDIA A100 with 8 GPUs for 5 days, has an approximate power consumption as follows:
 
 It's important to note that the actual power consumption may vary depending on the specific workload and operational conditions. For accurate power consumption measurements, using dedicated power monitoring tools is recommended.
 
 | Model          | GPU Type            | Power Consumption From GPUs | 
 |----------------|---------------------|-----------------------------|
-| GRAG-PHI-SFT   | A100 ([Hessian AI supercomputer](https://hessian.ai/de/)) | 0.288 MWh                    |
+| GRAG-NEMO-SFT   | A100 ([Hessian AI supercomputer](https://hessian.ai/de/)) | 0.288 MWh                    |
 ## Bias, Risks, and Limitations
 
 Like any base language model or fine-tuned model without safety filtering, it is relatively easy for a user to prompt these models to generate harmful and generally sensitive content.
 Such content can also be produced unintentionally, especially in the case of bias, so we recommend users consider the risks of applications of this technology.
 
-Otherwise, many facts from GRAG-Phi-SFT or any LLM will often not be true, so they should be checked.
+Otherwise, many facts from GRAG-NEMO-SFT or any LLM will often not be true, so they should be checked.