Commit
•
5577f74
1
Parent(s):
e7e5fc7
Update README.md
Browse files
README.md
CHANGED
|
@@ -57,6 +57,16 @@ The following hyperparameters were used during training:
|
|
| 57 |
|:-------------:|:------:|:----:|:---------------:|
|
| 58 |
| 1.2093 | 2.8429 | 500 | 1.7462 |
|
| 59 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 60 |
|
| 61 |
### Framework versions
|
| 62 |
|
|
|
|
| 57 |
|:-------------:|:------:|:----:|:---------------:|
|
| 58 |
| 1.2093 | 2.8429 | 500 | 1.7462 |
|
| 59 |
|
| 60 |
+
### Evaluation results
|
| 61 |
+
| Name | Checkpoint | Rouge1 | RougeL | Meteor | Bert Score |
|
| 62 |
+
|-------------------------|------------------------------------------------------------------------------------------------------------------|---------|--------|--------|------------|
|
| 63 |
+
| baseline instruct model | [Meta-Llama-3-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct) | 0.254 | 0.128 | 0.222 | 0.747 |
|
| 64 |
+
| full fientune from inst | [sft_llama3_instruct_full](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct) | 0.315 | 0.189 | 0.238 | 0.782 |
|
| 65 |
+
| lora sft from inst | [sft_llama3_instruct_lora_all](https://huggingface.co/geshijoker/HealthCareMagic_sft_llama3_instruct_lora_all) | 0.271 | 0.143 | 0.194 | 0.774 |
|
| 66 |
+
| lora sft from base | [sft_llama3_lora_all](geshijoker/HealthCareMagic_sft_llama3_lora_all) | 0.239 | 0.113 | 0.211 | 0.735 |
|
| 67 |
+
| qlora sft from inst | [sft_llama3_instruct_qlora_all](https://huggingface.co/geshijoker/HealthCareMagic_sft_llama3_instruct_qlora_all) | 0.137 | 0.071 | 0.102 | 0.679 |
|
| 68 |
+
| qlora sft from base | [sft_llama3_qlora_all ](https://huggingface.co/geshijoker/HealthCareMagic_sft_llama3_qlora_all) | 0.192 | 0.090 | 0.159 | 0.718 |
|
| 69 |
+
|
| 70 |
|
| 71 |
### Framework versions
|
| 72 |
|