Upload folder using huggingface_hub (#3)

- 7d93559dbfab971c4f9a076747f417f08d7288c904f5a7987624e6600b2c6b44 (cf7578d10a2f56348ef30ed0f3607010b0ee8d84)
- 7b56ee73980d90594b97f63a2d5ee97c571f3d2a63564c2c1f46f385721137de (cd7dabef2c8a10faeb98c836575684235f5178fd)

README.md

@@ -34,7 +34,7 @@ tags:
 
 ## Results
 
-Detailed efficiency metrics coming soon!
+![image info](./plots.png)
 
 **Frequently Asked Questions**
 - ***How does the compression work?*** The model is compressed with llm-int8.
@@ -61,7 +61,7 @@ You can run the smashed model with these steps:
    from transformers import AutoModelForCausalLM, AutoTokenizer
 
     model = AutoModelForCausalLM.from_pretrained("PrunaAI/lmsys-vicuna-7b-v1.3-bnb-4bit-smashed",
-                                                 trust_remote_code=True)
+                                                 trust_remote_code=True, device_map='auto')
     tokenizer = AutoTokenizer.from_pretrained("lmsys/vicuna-7b-v1.3")
     
     input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"]

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "/tmp/tmp9x3ypjfm",
+  "_name_or_path": "/tmp/tmp_5p953iv",
   "architectures": [
     "LlamaForCausalLM"
   ],
@@ -21,7 +21,7 @@
   "quantization_config": {
     "bnb_4bit_compute_dtype": "bfloat16",
     "bnb_4bit_quant_type": "fp4",
-    "bnb_4bit_use_double_quant": true,
+    "bnb_4bit_use_double_quant": false,
     "llm_int8_enable_fp32_cpu_offload": false,
     "llm_int8_has_fp16_weight": false,
     "llm_int8_skip_modules": [

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c5ddc3c2e2d48c4969b566db382ef553383b6cfdd16fcc3c71419b7b4c7f5a12
-size 3866042065
+oid sha256:9f05b2b415d3ba0af97e81163ef894ef677a606ebbb12186ca606ed4393ffc83
+size 4167722008

smash_config.json

@@ -8,7 +8,7 @@
         "compilers": "None",
         "task": "text_text_generation",
         "device": "cuda",
-        "cache_dir": "/ceph/hdd/staff/charpent/.cache/models1onx03oj",
+        "cache_dir": "/ceph/hdd/staff/charpent/.cache/modelsmlf3meh7",
         "batch_size": 1,
         "model_name": "lmsys/vicuna-7b-v1.3",
         "pruning_ratio": 0.0,