Upload folder using huggingface_hub (#3)

- 037b1835d9e00f7e35d3f1d8a51a88d6c9c935eaac8b7045bbd973ab2c485eb5 (ca550e321a3c8de0480fcc5478a8a2231ecdc358)
- f87dfc01ecef12fabda8c647ec6039698af0b9d6a739cf9e8ed2dbe33e8111c7 (66dc9314b7718a8f86f3b6999a738d3a4af3127a)

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/deepseek-ai-deepseek-coder-6.7b-instruct-bnb-4bit-smashed",
-                                                 trust_remote_code=True)
+                                                 trust_remote_code=True, device_map='auto')
     tokenizer = AutoTokenizer.from_pretrained("deepseek-ai/deepseek-coder-6.7b-instruct")
     
     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/tmpoj5rufer",
+  "_name_or_path": "/tmp/tmpqdorz7v6",
   "architectures": [
     "LlamaForCausalLM"
   ],
@@ -20,7 +20,7 @@
   "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 sha256:2a3794a34c8e3ca277a575a9f83806811bcf1fbe0160e7af551f079b9206812d
-size 3870236308
+oid sha256:8acf719c19eb932d2aef260a9a174022265f328dd52448a883d92c867f45e684
+size 4171916320

smash_config.json

@@ -8,7 +8,7 @@
         "compilers": "None",
         "task": "text_text_generation",
         "device": "cuda",
-        "cache_dir": "/ceph/hdd/staff/charpent/.cache/modelswlwxuu69",
+        "cache_dir": "/ceph/hdd/staff/charpent/.cache/modelsjhwn2bur",
         "batch_size": 1,
         "model_name": "deepseek-ai/deepseek-coder-6.7b-instruct",
         "pruning_ratio": 0.0,