Instructions to use Sapka/bloom-560m-customer with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use Sapka/bloom-560m-customer with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("bigscience/bloom-560m") model = PeftModel.from_pretrained(base_model, "Sapka/bloom-560m-customer") - Notebooks
- Google Colab
- Kaggle
basic training
Browse files
README.md
CHANGED
|
@@ -201,6 +201,17 @@ Carbon emissions can be estimated using the [Machine Learning Impact calculator]
|
|
| 201 |
## Training procedure
|
| 202 |
|
| 203 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 204 |
### Framework versions
|
| 205 |
|
| 206 |
|
|
|
|
| 201 |
## Training procedure
|
| 202 |
|
| 203 |
|
| 204 |
+
The following `bitsandbytes` quantization config was used during training:
|
| 205 |
+
- load_in_8bit: True
|
| 206 |
+
- load_in_4bit: False
|
| 207 |
+
- llm_int8_threshold: 6.0
|
| 208 |
+
- llm_int8_skip_modules: None
|
| 209 |
+
- llm_int8_enable_fp32_cpu_offload: False
|
| 210 |
+
- llm_int8_has_fp16_weight: False
|
| 211 |
+
- bnb_4bit_quant_type: fp4
|
| 212 |
+
- bnb_4bit_use_double_quant: False
|
| 213 |
+
- bnb_4bit_compute_dtype: float32
|
| 214 |
+
|
| 215 |
### Framework versions
|
| 216 |
|
| 217 |
|