Update README.md
Browse files
README.md
CHANGED
|
@@ -33,7 +33,7 @@ Without their independent research collaboration this model release would not ha
|
|
| 33 |
# Table of Contents
|
| 34 |
1. [Overview of all SauerkrautLM-7b-LaserChat models](#all-sauerkrautlm-7b-laserchat-models)
|
| 35 |
2. [Model Details](#model-details)
|
| 36 |
-
|
| 37 |
- [Prompt template](#prompt-template)
|
| 38 |
- [Training procedure](#proceed-of-the-training)
|
| 39 |
3. [Evaluation](#evaluation)
|
|
@@ -75,7 +75,9 @@ This process not only helps in understanding the effectiveness of Spherical Line
|
|
| 75 |
|
| 76 |
Additionally, we integrated a novel training strategy on the SFT and DPO training process, where we partially freeze the model according to a laser-like analysis aiming to navigate and optimize the trade-offs highlighted by the no free lunch theorem. This innovative training method effectively prevents the significant problem of language models forgetting previously acquired knowledge.
|
| 77 |
This aspect is particularly crucial when attempting to teach the model specific skills, such as a new language, where in general, the model might lose a considerable amount of its prior knowledge and exhibit a decline in overall intelligence.
|
|
|
|
| 78 |
**For function calling, we provide several branches with different versions of the model.Since Function Calling is currently still in beta status, we depend on your feedback. Please test each model extensively and let us know which model you achieved the best results with.**
|
|
|
|
| 79 |
Detailed information on how the new training strategy works and the advantages it offers over conventional training methods will soon be published in a detailed paper by the LaserRMT research group.
|
| 80 |
|
| 81 |
|
|
|
|
| 33 |
# Table of Contents
|
| 34 |
1. [Overview of all SauerkrautLM-7b-LaserChat models](#all-sauerkrautlm-7b-laserchat-models)
|
| 35 |
2. [Model Details](#model-details)
|
| 36 |
+
- [Function Calling Prompt template](#function-calling-prompt-template)
|
| 37 |
- [Prompt template](#prompt-template)
|
| 38 |
- [Training procedure](#proceed-of-the-training)
|
| 39 |
3. [Evaluation](#evaluation)
|
|
|
|
| 75 |
|
| 76 |
Additionally, we integrated a novel training strategy on the SFT and DPO training process, where we partially freeze the model according to a laser-like analysis aiming to navigate and optimize the trade-offs highlighted by the no free lunch theorem. This innovative training method effectively prevents the significant problem of language models forgetting previously acquired knowledge.
|
| 77 |
This aspect is particularly crucial when attempting to teach the model specific skills, such as a new language, where in general, the model might lose a considerable amount of its prior knowledge and exhibit a decline in overall intelligence.
|
| 78 |
+
|
| 79 |
**For function calling, we provide several branches with different versions of the model.Since Function Calling is currently still in beta status, we depend on your feedback. Please test each model extensively and let us know which model you achieved the best results with.**
|
| 80 |
+
|
| 81 |
Detailed information on how the new training strategy works and the advantages it offers over conventional training methods will soon be published in a detailed paper by the LaserRMT research group.
|
| 82 |
|
| 83 |
|