Upload FalconForCausalLM
Browse files- config.json +13 -1
config.json
CHANGED
|
@@ -28,7 +28,19 @@
|
|
| 28 |
"num_hidden_layers": 32,
|
| 29 |
"num_kv_heads": 71,
|
| 30 |
"parallel_attn": true,
|
| 31 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 32 |
"transformers_version": "4.37.0.dev0",
|
| 33 |
"use_cache": true,
|
| 34 |
"vocab_size": 65024
|
|
|
|
| 28 |
"num_hidden_layers": 32,
|
| 29 |
"num_kv_heads": 71,
|
| 30 |
"parallel_attn": true,
|
| 31 |
+
"quantization_config": {
|
| 32 |
+
"bnb_4bit_compute_dtype": "bfloat16",
|
| 33 |
+
"bnb_4bit_quant_type": "nf4",
|
| 34 |
+
"bnb_4bit_use_double_quant": true,
|
| 35 |
+
"llm_int8_enable_fp32_cpu_offload": false,
|
| 36 |
+
"llm_int8_has_fp16_weight": false,
|
| 37 |
+
"llm_int8_skip_modules": null,
|
| 38 |
+
"llm_int8_threshold": 6.0,
|
| 39 |
+
"load_in_4bit": true,
|
| 40 |
+
"load_in_8bit": false,
|
| 41 |
+
"quant_method": "bitsandbytes"
|
| 42 |
+
},
|
| 43 |
+
"torch_dtype": "float16",
|
| 44 |
"transformers_version": "4.37.0.dev0",
|
| 45 |
"use_cache": true,
|
| 46 |
"vocab_size": 65024
|