yhyu13
commited on
Commit
·
a1083e9
1
Parent(s):
7d20848
Update
Browse files
README.md
CHANGED
|
@@ -3,4 +3,80 @@ license: other
|
|
| 3 |
license_name: microsoft-research-license
|
| 4 |
license_link: >-
|
| 5 |
https://huggingface.co/cognitivecomputations/dolphin-2_6-phi-2/blob/main/LICENSE
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
license_name: microsoft-research-license
|
| 4 |
license_link: >-
|
| 5 |
https://huggingface.co/cognitivecomputations/dolphin-2_6-phi-2/blob/main/LICENSE
|
| 6 |
+
library_name: peft
|
| 7 |
+
tags:
|
| 8 |
+
- llama-factory
|
| 9 |
+
- lora
|
| 10 |
+
- generated_from_trainer
|
| 11 |
+
base_model: ./models/dolphin-2_6-phi-2
|
| 12 |
+
model-index:
|
| 13 |
+
- name: dolphin-2_6-phi-2-sft-glaive-function-calling-v2-ep1-lora
|
| 14 |
+
results: []
|
| 15 |
---
|
| 16 |
+
|
| 17 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 18 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 19 |
+
|
| 20 |
+
# dolphin-2_6-phi-2-sft-glaive-function-calling-v2-ep1-lora
|
| 21 |
+
|
| 22 |
+
This model is a fine-tuned version of [./models/dolphin-2_6-phi-2](https://huggingface.co/./models/dolphin-2_6-phi-2) on the simple-function-calling-v2_convert dataset that I converted for llama_factory https://huggingface.co/datasets/Yhyu13/glaive-function-calling-v2-llama-factory-convert.
|
| 23 |
+
It achieves the following results on the evaluation set:
|
| 24 |
+
- Loss: 0.3524
|
| 25 |
+
|
| 26 |
+
## Model description
|
| 27 |
+
|
| 28 |
+
More information needed
|
| 29 |
+
|
| 30 |
+
## Intended uses & limitations
|
| 31 |
+
|
| 32 |
+
More information needed
|
| 33 |
+
|
| 34 |
+
## Training and evaluation data
|
| 35 |
+
|
| 36 |
+
More information needed
|
| 37 |
+
|
| 38 |
+
## Training procedure
|
| 39 |
+
|
| 40 |
+
|
| 41 |
+
The following `bitsandbytes` quantization config was used during training:
|
| 42 |
+
- quant_method: QuantizationMethod.BITS_AND_BYTES
|
| 43 |
+
- load_in_8bit: False
|
| 44 |
+
- load_in_4bit: True
|
| 45 |
+
- llm_int8_threshold: 6.0
|
| 46 |
+
- llm_int8_skip_modules: None
|
| 47 |
+
- llm_int8_enable_fp32_cpu_offload: False
|
| 48 |
+
- llm_int8_has_fp16_weight: False
|
| 49 |
+
- bnb_4bit_quant_type: nf4
|
| 50 |
+
- bnb_4bit_use_double_quant: True
|
| 51 |
+
- bnb_4bit_compute_dtype: float16
|
| 52 |
+
|
| 53 |
+
### Training hyperparameters
|
| 54 |
+
|
| 55 |
+
The following hyperparameters were used during training:
|
| 56 |
+
- learning_rate: 5e-05
|
| 57 |
+
- train_batch_size: 1
|
| 58 |
+
- eval_batch_size: 1
|
| 59 |
+
- seed: 42
|
| 60 |
+
- distributed_type: multi-GPU
|
| 61 |
+
- num_devices: 2
|
| 62 |
+
- gradient_accumulation_steps: 4
|
| 63 |
+
- total_train_batch_size: 8
|
| 64 |
+
- total_eval_batch_size: 2
|
| 65 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 66 |
+
- lr_scheduler_type: cosine
|
| 67 |
+
- num_epochs: 1.0
|
| 68 |
+
|
| 69 |
+
### Training results
|
| 70 |
+
|
| 71 |
+
| Training Loss | Epoch | Step | Validation Loss |
|
| 72 |
+
|:-------------:|:-----:|:----:|:---------------:|
|
| 73 |
+
| 0.3453 | 1.0 | 376 | 0.3524 |
|
| 74 |
+
|
| 75 |
+
|
| 76 |
+
### Framework versions
|
| 77 |
+
|
| 78 |
+
- PEFT 0.7.0
|
| 79 |
+
- Transformers 4.36.2
|
| 80 |
+
- Pytorch 2.1.1+cu121
|
| 81 |
+
- Datasets 2.14.7
|
| 82 |
+
- Tokenizers 0.15.0
|