Upload folder using huggingface_hub (#3)

- f8f37e7767f87867a7472cf8e6b33558d0fe55ae4bc0106f3c100fdfadcbaaa4 (a63fef142cf79a1081fe22998f137070f1abfb58)
- 5423df0d295b379501e2130cfb35920c27a899ced635f87aecdac35f42747337 (67f3f1094d27095f5a2aeee23455cd03aa2090b6)

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/beomi-OPEN-SOLAR-KO-10.7B-bnb-4bit-smashed",
-                                                 trust_remote_code=True)
+                                                 trust_remote_code=True, device_map='auto')
     tokenizer = AutoTokenizer.from_pretrained("beomi/OPEN-SOLAR-KO-10.7B")
     
     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/tmp1mt56sgx",
+  "_name_or_path": "/tmp/tmp6sjg8w72",
   "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-00001-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f7ede0e916b5cbe0172ded557cd5b7474b3d1f64ac1c465637cb47a3ffed6eef
-size 4995964476
+oid sha256:e54ff3d3d4b907042299113ed3dc1a1ad8fdf8927f9ee13d342ac7a51b4bd79c
+size 4978371160

model-00002-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f6fb9e32dad6f28290ad697d5a188b7e159c40a6b0f56e0e7c8ba42bf1f53fb9
-size 1169494101
+oid sha256:c99b5f6da860b29ed810ed4bffd6112b4fa981718e3711af0825dae9e26310e7
+size 1674812136

smash_config.json

@@ -8,7 +8,7 @@
         "compilers": "None",
         "task": "text_text_generation",
         "device": "cuda",
-        "cache_dir": "/ceph/hdd/staff/charpent/.cache/models8o8o6741",
+        "cache_dir": "/ceph/hdd/staff/charpent/.cache/modelswzjl4uav",
         "batch_size": 1,
         "model_name": "beomi/OPEN-SOLAR-KO-10.7B",
         "pruning_ratio": 0.0,