Upload folder using huggingface_hub (#2)

- 5adb3a1b71fa379b842b50838663dd449b68f1b2724c6d5458279f750e361549 (0984dcc7c6a41f6564bd71a96169f39f36bf80e6)
- 92e563b6d4fb58b80e75c8496c800cf31623658f05c276c6a6e58c6e4b3954dd (da3affc028ca76a2aff9386f31a3c5fdee66f807)
- dec34f453b170eb247344c1b08e80ef4084cf65e90e43f1f1826fc5cdbefd77d (faba919aeb856ddef458a562e6458254462b6c2b)

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/Intel-neural-chat-7b-v3-3-bnb-8bit-smashed",
-                                                 trust_remote_code=True)
+                                                 trust_remote_code=True, device_map='auto')
     tokenizer = AutoTokenizer.from_pretrained("Intel/neural-chat-7b-v3-3")
     
     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/tmpxjnbn5b3",
+  "_name_or_path": "/tmp/tmp3x3nt9v8",
   "architectures": [
     "MistralForCausalLM"
   ],
@@ -18,7 +18,7 @@
   "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": [

smash_config.json

@@ -8,7 +8,7 @@
         "compilers": "None",
         "task": "text_text_generation",
         "device": "cuda",
-        "cache_dir": "/ceph/hdd/staff/charpent/.cache/models8ibp523l",
+        "cache_dir": "/ceph/hdd/staff/charpent/.cache/models_7b_0r8h",
         "batch_size": 1,
         "model_name": "Intel/neural-chat-7b-v3-3",
         "pruning_ratio": 0.0,