Upload folder using huggingface_hub
Browse files- README.md +23 -0
- config.json +51 -0
- model.safetensors +3 -0
README.md
ADDED
|
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model:
|
| 3 |
+
- medmekk/Llama-3.2-1B-BNB-int4-nf4
|
| 4 |
+
---
|
| 5 |
+
|
| 6 |
+
# medmekk/Llama-3.2-1B-BNB-int4-nf4 (Quantized)
|
| 7 |
+
|
| 8 |
+
## Description
|
| 9 |
+
This model is a quantized version of the original model `medmekk/Llama-3.2-1B-BNB-int4-nf4`. It has been quantized using int4 quantization with bitsandbytes.
|
| 10 |
+
|
| 11 |
+
## Quantization Details
|
| 12 |
+
- **Quantization Type**: int4
|
| 13 |
+
- **Threshold**: None
|
| 14 |
+
- **bnb_4bit_quant_type**: nf4
|
| 15 |
+
- **bnb_4bit_use_double_quant**: False
|
| 16 |
+
|
| 17 |
+
## Usage
|
| 18 |
+
You can use this model in your applications by loading it directly from the Hugging Face Hub:
|
| 19 |
+
|
| 20 |
+
```python
|
| 21 |
+
from transformers import AutoModel
|
| 22 |
+
|
| 23 |
+
model = AutoModel.from_pretrained("medmekk/Llama-3.2-1B-BNB-int4-nf4")
|
config.json
ADDED
|
@@ -0,0 +1,51 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "meta-llama/Llama-3.2-1B",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"LlamaModel"
|
| 5 |
+
],
|
| 6 |
+
"attention_bias": false,
|
| 7 |
+
"attention_dropout": 0.0,
|
| 8 |
+
"bos_token_id": 128000,
|
| 9 |
+
"eos_token_id": 128001,
|
| 10 |
+
"head_dim": 64,
|
| 11 |
+
"hidden_act": "silu",
|
| 12 |
+
"hidden_size": 2048,
|
| 13 |
+
"initializer_range": 0.02,
|
| 14 |
+
"intermediate_size": 8192,
|
| 15 |
+
"max_position_embeddings": 131072,
|
| 16 |
+
"mlp_bias": false,
|
| 17 |
+
"model_type": "llama",
|
| 18 |
+
"num_attention_heads": 32,
|
| 19 |
+
"num_hidden_layers": 16,
|
| 20 |
+
"num_key_value_heads": 8,
|
| 21 |
+
"pretraining_tp": 1,
|
| 22 |
+
"quantization_config": {
|
| 23 |
+
"_load_in_4bit": true,
|
| 24 |
+
"_load_in_8bit": false,
|
| 25 |
+
"bnb_4bit_compute_dtype": "float32",
|
| 26 |
+
"bnb_4bit_quant_storage": "uint8",
|
| 27 |
+
"bnb_4bit_quant_type": "nf4",
|
| 28 |
+
"bnb_4bit_use_double_quant": false,
|
| 29 |
+
"llm_int8_enable_fp32_cpu_offload": false,
|
| 30 |
+
"llm_int8_has_fp16_weight": false,
|
| 31 |
+
"llm_int8_skip_modules": null,
|
| 32 |
+
"llm_int8_threshold": 6.0,
|
| 33 |
+
"load_in_4bit": true,
|
| 34 |
+
"load_in_8bit": false,
|
| 35 |
+
"quant_method": "bitsandbytes"
|
| 36 |
+
},
|
| 37 |
+
"rms_norm_eps": 1e-05,
|
| 38 |
+
"rope_scaling": {
|
| 39 |
+
"factor": 32.0,
|
| 40 |
+
"high_freq_factor": 4.0,
|
| 41 |
+
"low_freq_factor": 1.0,
|
| 42 |
+
"original_max_position_embeddings": 8192,
|
| 43 |
+
"rope_type": "llama3"
|
| 44 |
+
},
|
| 45 |
+
"rope_theta": 500000.0,
|
| 46 |
+
"tie_word_embeddings": true,
|
| 47 |
+
"torch_dtype": "float16",
|
| 48 |
+
"transformers_version": "4.49.0.dev0",
|
| 49 |
+
"use_cache": true,
|
| 50 |
+
"vocab_size": 128256
|
| 51 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:91583b1659ef877fe422422113771ec6ab7a42be64233fda12e66656d21f7967
|
| 3 |
+
size 1072898712
|