070713efb3ab710778692e3cd66a8363629f201c90c58c9a388d2d04d611f8ee

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/42dot-42dot_LLM-SFT-1.3B-bnb-4bit-smashed",
-                                                 trust_remote_code=True)
+                                                 trust_remote_code=True, device_map='auto')
     tokenizer = AutoTokenizer.from_pretrained("42dot/42dot_LLM-SFT-1.3B")
     
     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/tmpcy1rf9t7",
+  "_name_or_path": "/tmp/tmpkwdgp3ya",
   "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:f2b28cabf52f8bc8d4109d42f50db805ec7a7ea15909de5ba4227d66e322dd5b
-size 1048758241
+oid sha256:d3f4fc05455018bbc584f6ce59f233bfbd80b45a94a59473b2b7c28c34417d8b
+size 1106034112

smash_config.json

@@ -8,7 +8,7 @@
         "compilers": "None",
         "task": "text_text_generation",
         "device": "cuda",
-        "cache_dir": "/ceph/hdd/staff/charpent/.cache/modelsxe52cy0p",
+        "cache_dir": "/ceph/hdd/staff/charpent/.cache/modelsw534s7vh",
         "batch_size": 1,
         "model_name": "42dot/42dot_LLM-SFT-1.3B",
         "pruning_ratio": 0.0,