Upload model
Browse files
README.md
CHANGED
|
@@ -1,7 +1,8 @@
|
|
| 1 |
---
|
| 2 |
-
|
| 3 |
tags:
|
| 4 |
- generated_from_trainer
|
|
|
|
| 5 |
model-index:
|
| 6 |
- name: outputs
|
| 7 |
results: []
|
|
@@ -53,3 +54,22 @@ The following hyperparameters were used during training:
|
|
| 53 |
- Pytorch 2.1.0+cu118
|
| 54 |
- Datasets 2.15.0
|
| 55 |
- Tokenizers 0.15.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
+
library_name: peft
|
| 3 |
tags:
|
| 4 |
- generated_from_trainer
|
| 5 |
+
base_model: NousResearch/Llama-2-7b-chat-hf
|
| 6 |
model-index:
|
| 7 |
- name: outputs
|
| 8 |
results: []
|
|
|
|
| 54 |
- Pytorch 2.1.0+cu118
|
| 55 |
- Datasets 2.15.0
|
| 56 |
- Tokenizers 0.15.0
|
| 57 |
+
## Training procedure
|
| 58 |
+
|
| 59 |
+
|
| 60 |
+
The following `bitsandbytes` quantization config was used during training:
|
| 61 |
+
- quant_method: bitsandbytes
|
| 62 |
+
- load_in_8bit: False
|
| 63 |
+
- load_in_4bit: True
|
| 64 |
+
- llm_int8_threshold: 6.0
|
| 65 |
+
- llm_int8_skip_modules: None
|
| 66 |
+
- llm_int8_enable_fp32_cpu_offload: False
|
| 67 |
+
- llm_int8_has_fp16_weight: False
|
| 68 |
+
- bnb_4bit_quant_type: nf4
|
| 69 |
+
- bnb_4bit_use_double_quant: True
|
| 70 |
+
- bnb_4bit_compute_dtype: bfloat16
|
| 71 |
+
|
| 72 |
+
### Framework versions
|
| 73 |
+
|
| 74 |
+
|
| 75 |
+
- PEFT 0.6.3.dev0
|