Hugging Face's logo

behzadnet
/
Llama-2-7b-chat-hf-sharded-bf16-fine-tuned-adapters_GroundTruth_lr4_seed102

PEFT
Model card Files Community
behzadnet commited on
Commit
a436446
·
1 Parent(s): 11c8830

Upload model

Browse files
Files changed (3) hide show
  1. README.md +72 -0
  2. adapter_config.json +3 -3
  3. adapter_model.bin +1 -1
README.md CHANGED
@@ -233,4 +233,76 @@ The following `bitsandbytes` quantization config was used during training:
233
  ### Framework versions
234
 
235
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
236
  - PEFT 0.7.0.dev0
 
233
  ### Framework versions
234
 
235
 
236
+ - PEFT 0.7.0.dev0
237
+ ## Training procedure
238
+
239
+
240
+ The following `bitsandbytes` quantization config was used during training:
241
+ - load_in_8bit: False
242
+ - load_in_4bit: True
243
+ - llm_int8_threshold: 6.0
244
+ - llm_int8_skip_modules: None
245
+ - llm_int8_enable_fp32_cpu_offload: False
246
+ - llm_int8_has_fp16_weight: False
247
+ - bnb_4bit_quant_type: nf4
248
+ - bnb_4bit_use_double_quant: True
249
+ - bnb_4bit_compute_dtype: bfloat16
250
+
251
+ ### Framework versions
252
+
253
+
254
+ - PEFT 0.7.0.dev0
255
+ ## Training procedure
256
+
257
+
258
+ The following `bitsandbytes` quantization config was used during training:
259
+ - load_in_8bit: False
260
+ - load_in_4bit: True
261
+ - llm_int8_threshold: 6.0
262
+ - llm_int8_skip_modules: None
263
+ - llm_int8_enable_fp32_cpu_offload: False
264
+ - llm_int8_has_fp16_weight: False
265
+ - bnb_4bit_quant_type: nf4
266
+ - bnb_4bit_use_double_quant: True
267
+ - bnb_4bit_compute_dtype: bfloat16
268
+
269
+ ### Framework versions
270
+
271
+
272
+ - PEFT 0.7.0.dev0
273
+ ## Training procedure
274
+
275
+
276
+ The following `bitsandbytes` quantization config was used during training:
277
+ - load_in_8bit: False
278
+ - load_in_4bit: True
279
+ - llm_int8_threshold: 6.0
280
+ - llm_int8_skip_modules: None
281
+ - llm_int8_enable_fp32_cpu_offload: False
282
+ - llm_int8_has_fp16_weight: False
283
+ - bnb_4bit_quant_type: nf4
284
+ - bnb_4bit_use_double_quant: True
285
+ - bnb_4bit_compute_dtype: bfloat16
286
+
287
+ ### Framework versions
288
+
289
+
290
+ - PEFT 0.7.0.dev0
291
+ ## Training procedure
292
+
293
+
294
+ The following `bitsandbytes` quantization config was used during training:
295
+ - load_in_8bit: False
296
+ - load_in_4bit: True
297
+ - llm_int8_threshold: 6.0
298
+ - llm_int8_skip_modules: None
299
+ - llm_int8_enable_fp32_cpu_offload: False
300
+ - llm_int8_has_fp16_weight: False
301
+ - bnb_4bit_quant_type: nf4
302
+ - bnb_4bit_use_double_quant: True
303
+ - bnb_4bit_compute_dtype: bfloat16
304
+
305
+ ### Framework versions
306
+
307
+
308
  - PEFT 0.7.0.dev0
adapter_config.json CHANGED
@@ -16,10 +16,10 @@
16
  "rank_pattern": {},
17
  "revision": null,
18
  "target_modules": [
19
- "self_attn.q_proj",
20
  "self_attn.k_proj",
21
- "self_attn.o_proj",
22
- "self_attn.v_proj"
 
23
  ],
24
  "task_type": "CAUSAL_LM"
25
  }
 
16
  "rank_pattern": {},
17
  "revision": null,
18
  "target_modules": [
 
19
  "self_attn.k_proj",
20
+ "self_attn.q_proj",
21
+ "self_attn.v_proj",
22
+ "self_attn.o_proj"
23
  ],
24
  "task_type": "CAUSAL_LM"
25
  }
adapter_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b6dfc300a3a19dc7b8538cae347b58e4f45e1d9f80f83abea9fcd8be76fd0a35
3
  size 33646413
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:69986dda42452b5dfc0bc1d22f67db151422af08ce52471674e5284647c74319
3
  size 33646413