cee4dc36fb8258e628fd49c0fbc72967d96565d26f8f888ae0055a203428126f

README.md

@@ -1,5 +1,4 @@
 ---
-library_name: pruna-engine
 thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
 metrics:
 - memory_disk
@@ -8,6 +7,8 @@ metrics:
 - inference_throughput
 - inference_CO2_emissions
 - inference_energy_consumption
+tags:
+- pruna-ai
 ---
 <!-- header start -->
 <!-- 200823 -->
@@ -33,7 +34,7 @@ metrics:
 
 ## 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.
@@ -60,7 +61,7 @@ You can run the smashed model with these steps:
    from transformers import AutoModelForCausalLM, AutoTokenizer
 
     model = AutoModelForCausalLM.from_pretrained("PrunaAI/princeton-nlp-Sheared-LLaMA-2.7B-bnb-4bit-smashed",
-                                                 trust_remote_code=True)
+                                                 trust_remote_code=True, device_map='auto')
     tokenizer = AutoTokenizer.from_pretrained("princeton-nlp/Sheared-LLaMA-2.7B")
     
     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/tmped1mjtnh",
+  "_name_or_path": "/tmp/tmpmvrdgwvz",
   "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 sha256:65302a2039b112a4b563f9e5c335b6068e5ae4af9386d967a3fd2374bf74bc5a
-size 1637507272
+oid sha256:298957571aa412227b1aabb676b8aecbe73b2674fa67b2a5658a2d44bd146069
+size 1755534464

smash_config.json

@@ -8,7 +8,7 @@
         "compilers": "None",
         "task": "text_text_generation",
         "device": "cuda",
-        "cache_dir": "/ceph/hdd/staff/charpent/.cache/modelsdrrq2fuj",
+        "cache_dir": "/ceph/hdd/staff/charpent/.cache/modelsjnmaigha",
         "batch_size": 1,
         "model_name": "princeton-nlp/Sheared-LLaMA-2.7B",
         "pruning_ratio": 0.0,