Upload new GPTQs with varied parameters
Browse files
README.md
CHANGED
|
@@ -1,6 +1,9 @@
|
|
| 1 |
---
|
|
|
|
|
|
|
| 2 |
inference: false
|
| 3 |
license: other
|
|
|
|
| 4 |
---
|
| 5 |
|
| 6 |
<!-- header start -->
|
|
@@ -38,7 +41,6 @@ A chat between a curious user and an artificial intelligence assistant. The assi
|
|
| 38 |
|
| 39 |
USER: {prompt}
|
| 40 |
ASSISTANT:
|
| 41 |
-
|
| 42 |
```
|
| 43 |
|
| 44 |
## Provided files
|
|
@@ -51,8 +53,8 @@ Each separate quant is in a different branch. See below for instructions on fet
|
|
| 51 |
| ------ | ---- | ---------- | -------------------- | --------- | ------------------- | --------- | ----------- |
|
| 52 |
| main | 4 | 128 | False | 7.45 GB | True | GPTQ-for-LLaMa | Most compatible option. Good inference speed in AutoGPTQ and GPTQ-for-LLaMa. Lower inference quality than other options. |
|
| 53 |
| gptq-4bit-32g-actorder_True | 4 | 32 | True | 8.00 GB | True | AutoGPTQ | 4-bit, with Act Order and group size. 32g gives highest possible inference quality, with maximum VRAM usage. Poor AutoGPTQ CUDA speed. |
|
| 54 |
-
| gptq-4bit-64g-actorder_True | 4 | 64 | True | 7.51 GB | True | AutoGPTQ | 4-bit, with Act Order and group size. 64g uses less VRAM, but with slightly lower accuracy. Poor AutoGPTQ CUDA speed. |
|
| 55 |
-
| gptq-4bit-128g-actorder_True | 4 | 128 | True | 7.26 GB | True | AutoGPTQ | 4-bit, with Act Order
|
| 56 |
| gptq-8bit--1g-actorder_True | 8 | None | True | 13.36 GB | False | AutoGPTQ | 8-bit, with Act Order. No group size, to lower VRAM requirements and to improve AutoGPTQ speed. |
|
| 57 |
| gptq-8bit-128g-actorder_False | 8 | 128 | False | 13.65 GB | False | AutoGPTQ | 8-bit, with group size 128g for higher inference quality and without Act Order to improve AutoGPTQ speed. |
|
| 58 |
|
|
@@ -128,7 +130,6 @@ prompt_template=f'''A chat between a curious user and an artificial intelligence
|
|
| 128 |
|
| 129 |
USER: {prompt}
|
| 130 |
ASSISTANT:
|
| 131 |
-
|
| 132 |
'''
|
| 133 |
|
| 134 |
print("\n\n*** Generate:")
|
|
|
|
| 1 |
---
|
| 2 |
+
datasets:
|
| 3 |
+
- stingning/ultrachat
|
| 4 |
inference: false
|
| 5 |
license: other
|
| 6 |
+
model_type: llama
|
| 7 |
---
|
| 8 |
|
| 9 |
<!-- header start -->
|
|
|
|
| 41 |
|
| 42 |
USER: {prompt}
|
| 43 |
ASSISTANT:
|
|
|
|
| 44 |
```
|
| 45 |
|
| 46 |
## Provided files
|
|
|
|
| 53 |
| ------ | ---- | ---------- | -------------------- | --------- | ------------------- | --------- | ----------- |
|
| 54 |
| main | 4 | 128 | False | 7.45 GB | True | GPTQ-for-LLaMa | Most compatible option. Good inference speed in AutoGPTQ and GPTQ-for-LLaMa. Lower inference quality than other options. |
|
| 55 |
| gptq-4bit-32g-actorder_True | 4 | 32 | True | 8.00 GB | True | AutoGPTQ | 4-bit, with Act Order and group size. 32g gives highest possible inference quality, with maximum VRAM usage. Poor AutoGPTQ CUDA speed. |
|
| 56 |
+
| gptq-4bit-64g-actorder_True | 4 | 64 | True | 7.51 GB | True | AutoGPTQ | 4-bit, with Act Order and group size. 64g uses less VRAM than 32g, but with slightly lower accuracy. Poor AutoGPTQ CUDA speed. |
|
| 57 |
+
| gptq-4bit-128g-actorder_True | 4 | 128 | True | 7.26 GB | True | AutoGPTQ | 4-bit, with Act Order and group size. 128g uses even less VRAM, but with slightly lower accuracy. Poor AutoGPTQ CUDA speed. |
|
| 58 |
| gptq-8bit--1g-actorder_True | 8 | None | True | 13.36 GB | False | AutoGPTQ | 8-bit, with Act Order. No group size, to lower VRAM requirements and to improve AutoGPTQ speed. |
|
| 59 |
| gptq-8bit-128g-actorder_False | 8 | 128 | False | 13.65 GB | False | AutoGPTQ | 8-bit, with group size 128g for higher inference quality and without Act Order to improve AutoGPTQ speed. |
|
| 60 |
|
|
|
|
| 130 |
|
| 131 |
USER: {prompt}
|
| 132 |
ASSISTANT:
|
|
|
|
| 133 |
'''
|
| 134 |
|
| 135 |
print("\n\n*** Generate:")
|