Files changed (4) hide show
  1. .gitattributes +1 -0
  2. README.md +8 -11
  3. imgs/finch.jpg +0 -0
  4. pytorch_model.bin +2 -2
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ pytorch_model.bin filter=lfs diff=lfs merge=lfs -text
README.md CHANGED
@@ -1,6 +1,3 @@
1
- ---
2
- license: apache-2.0
3
- ---
4
  ### Huggingface RWKV Finch 14B Model
5
 
6
  > HF compatible model for Finch-14B.
@@ -26,7 +23,7 @@ tokenizer = AutoTokenizer.from_pretrained("RWKV/v6-Finch-14B-HF", trust_remote_c
26
 
27
  The following demonstrates the improvements from Eagle 7B to Finch 14B
28
 
29
- | | [Eagle 7B](https://huggingface.co/RWKV/v6-Finch-14B-HF) | [Finch 7B](https://huggingface.co/RWKV/v6-Finch-7B-HF) | [Finch 14B](https://huggingface.co/RWKV/v6-Finch-14B-HF) |
30
  | --- | --- | --- | --- |
31
  | [ARC](https://github.com/EleutherAI/lm-evaluation-harness/tree/main/lm_eval/tasks/arc) | 39.59% | 41.47% | 46.33% |
32
  | [HellaSwag](https://github.com/EleutherAI/lm-evaluation-harness/tree/main/lm_eval/tasks/hellaswag) | 53.09% | 55.96% | 57.69% |
@@ -59,8 +56,8 @@ User: {instruction}
59
  Assistant:"""
60
 
61
 
62
- model = AutoModelForCausalLM.from_pretrained("RWKV/v6-Finch-14B-HF", trust_remote_code=True).to(torch.float32)
63
- tokenizer = AutoTokenizer.from_pretrained("RWKV/v6-Finch-14B-HF", trust_remote_code=True)
64
 
65
  text = "请介绍北京的旅游景点"
66
  prompt = generate_prompt(text)
@@ -115,8 +112,8 @@ User: {instruction}
115
  Assistant:"""
116
 
117
 
118
- model = AutoModelForCausalLM.from_pretrained("RWKV/v6-Finch-14B-HF", trust_remote_code=True, torch_dtype=torch.float16).to(0)
119
- tokenizer = AutoTokenizer.from_pretrained("RWKV/v6-Finch-14B-HF", trust_remote_code=True)
120
 
121
  text = "介绍一下大熊猫"
122
  prompt = generate_prompt(text)
@@ -162,8 +159,8 @@ User: {instruction}
162
 
163
  Assistant:"""
164
 
165
- model = AutoModelForCausalLM.from_pretrained("RWKV/v6-Finch-14B-HF", trust_remote_code=True).to(torch.float32)
166
- tokenizer = AutoTokenizer.from_pretrained("RWKV/v6-Finch-14B-HF", trust_remote_code=True)
167
 
168
  texts = ["请介绍北京的旅游景点", "介绍一下大熊猫", "乌兰察布"]
169
  prompts = [generate_prompt(text) for text in texts]
@@ -214,6 +211,6 @@ Assistant: 乌兰察布是中国新疆维吾尔自治区的一个县级市,位
214
  ## Acknowledgement
215
  We are grateful for the help and support from the following key groups:
216
 
217
- - [Recursal.ai](https://recursal.ai) team for financing the GPU resources, and managing the training of this foundation model - you can run the Finch line of RWKV models on their cloud / on-premise platform today.
218
  - EleutherAI for their support, especially in the v5/v6 Eagle/Finch paper
219
  - Linux Foundation AI & Data group for supporting and hosting the RWKV project
 
 
 
 
1
  ### Huggingface RWKV Finch 14B Model
2
 
3
  > HF compatible model for Finch-14B.
 
23
 
24
  The following demonstrates the improvements from Eagle 7B to Finch 14B
25
 
26
+ | | [Eagle 7B](https://huggingface.co/RWKV/v5-Eagle-7B-HF) | [Finch 7B](https://huggingface.co/RWKV/v6-Finch-7B-HF) | [Finch 14B](https://huggingface.co/RWKV/v6-Finch-14B-HF) |
27
  | --- | --- | --- | --- |
28
  | [ARC](https://github.com/EleutherAI/lm-evaluation-harness/tree/main/lm_eval/tasks/arc) | 39.59% | 41.47% | 46.33% |
29
  | [HellaSwag](https://github.com/EleutherAI/lm-evaluation-harness/tree/main/lm_eval/tasks/hellaswag) | 53.09% | 55.96% | 57.69% |
 
56
  Assistant:"""
57
 
58
 
59
+ model = AutoModelForCausalLM.from_pretrained("RWKV/v5-Eagle-7B-HF", trust_remote_code=True).to(torch.float32)
60
+ tokenizer = AutoTokenizer.from_pretrained("RWKV/v5-Eagle-7B-HF", trust_remote_code=True)
61
 
62
  text = "请介绍北京的旅游景点"
63
  prompt = generate_prompt(text)
 
112
  Assistant:"""
113
 
114
 
115
+ model = AutoModelForCausalLM.from_pretrained("RWKV/v5-Eagle-7B-HF", trust_remote_code=True, torch_dtype=torch.float16).to(0)
116
+ tokenizer = AutoTokenizer.from_pretrained("RWKV/v5-Eagle-7B-HF", trust_remote_code=True)
117
 
118
  text = "介绍一下大熊猫"
119
  prompt = generate_prompt(text)
 
159
 
160
  Assistant:"""
161
 
162
+ model = AutoModelForCausalLM.from_pretrained("RWKV/v5-Eagle-7B-HF", trust_remote_code=True).to(torch.float32)
163
+ tokenizer = AutoTokenizer.from_pretrained("RWKV/v5-Eagle-7B-HF", trust_remote_code=True)
164
 
165
  texts = ["请介绍北京的旅游景点", "介绍一下大熊猫", "乌兰察布"]
166
  prompts = [generate_prompt(text) for text in texts]
 
211
  ## Acknowledgement
212
  We are grateful for the help and support from the following key groups:
213
 
214
+ - [Recursal.ai](https://recursal.ai) team for financing the GPU resources, and managing the training of this foundation model - you can run the Eagle line of RWKV models on their cloud / on-premise platform today.
215
  - EleutherAI for their support, especially in the v5/v6 Eagle/Finch paper
216
  - Linux Foundation AI & Data group for supporting and hosting the RWKV project
imgs/finch.jpg DELETED
Binary file (315 kB)
 
pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f5f3477146b7dc0e68a1e64faac1e3b87a257d10f69a0078908d8d9bce6aaae9
3
- size 28138382794
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d8fd6a260cf910c13729c34109b47c4f28ec009764c79a25f8b35cf853997415
3
+ size 28138382858