Update README.md
Browse files
README.md
CHANGED
|
@@ -24,12 +24,16 @@ model-index:
|
|
| 24 |
metrics:
|
| 25 |
- name: pass@1
|
| 26 |
type: pass@1
|
| 27 |
-
value:
|
| 28 |
verified: false
|
| 29 |
---
|
| 30 |
|
| 31 |
<p><h1> speechless-code-mistral-7b-v1.0 </h1></p>
|
| 32 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
Use the following dataset to fine-tune mistralai/Mistral-7B-v0.1 in order to improve the model's reasoning and planning abilities.
|
| 34 |
|
| 35 |
Total 201,981 samples.
|
|
@@ -41,6 +45,40 @@ Total 201,981 samples.
|
|
| 41 |
- Spider: 8,659 samples
|
| 42 |
|
| 43 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 44 |
| | |
|
| 45 |
|------ | ------ |
|
| 46 |
| lr | 2e-4 |
|
|
@@ -75,32 +113,3 @@ A40-48G x 2
|
|
| 75 |
| eeval_runtime | 0:00:25.04 |
|
| 76 |
| eeval_samples_per_second | 7.985 |
|
| 77 |
| eeval_steps_per_second | |
|
| 78 |
-
|
| 79 |
-
| Metric | Value |
|
| 80 |
-
| --- | --- |
|
| 81 |
-
| humaneval-python ||
|
| 82 |
-
|
| 83 |
-
[Big Code Models Leaderboard](https://huggingface.co/spaces/bigcode/bigcode-models-leaderboard)
|
| 84 |
-
|
| 85 |
-
CodeLlama-34B-Python: 53.29
|
| 86 |
-
|
| 87 |
-
CodeLlama-34B-Instruct: 50.79
|
| 88 |
-
|
| 89 |
-
CodeLlama-13B-Instruct: 50.6
|
| 90 |
-
|
| 91 |
-
CodeLlama-34B: 45.11
|
| 92 |
-
|
| 93 |
-
CodeLlama-13B-Python: 42.89
|
| 94 |
-
|
| 95 |
-
CodeLlama-13B: 35.07
|
| 96 |
-
|
| 97 |
-
[Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
|
| 98 |
-
| Metric | Value |
|
| 99 |
-
| --- | --- |
|
| 100 |
-
| ARC | |
|
| 101 |
-
| HellaSwag | |
|
| 102 |
-
| MMLU | |
|
| 103 |
-
| TruthfulQA | |
|
| 104 |
-
| Average | |
|
| 105 |
-
|
| 106 |
-
|
|
|
|
| 24 |
metrics:
|
| 25 |
- name: pass@1
|
| 26 |
type: pass@1
|
| 27 |
+
value: 50.0
|
| 28 |
verified: false
|
| 29 |
---
|
| 30 |
|
| 31 |
<p><h1> speechless-code-mistral-7b-v1.0 </h1></p>
|
| 32 |
|
| 33 |
+
* [AWQ model(s) for GPU inference.](https://huggingface.co/TheBloke/speechless-code-mistral-7B-v1.0-AWQ)
|
| 34 |
+
* [GPTQ models for GPU inference, with multiple quantisation parameter options.](https://huggingface.co/TheBloke/speechless-code-mistral-7B-v1.0-GPTQ)
|
| 35 |
+
* [2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference](https://huggingface.co/TheBloke/speechless-code-mistral-7B-v1.0-GGUF)
|
| 36 |
+
|
| 37 |
Use the following dataset to fine-tune mistralai/Mistral-7B-v0.1 in order to improve the model's reasoning and planning abilities.
|
| 38 |
|
| 39 |
Total 201,981 samples.
|
|
|
|
| 45 |
- Spider: 8,659 samples
|
| 46 |
|
| 47 |
|
| 48 |
+
|
| 49 |
+
## HumanEval
|
| 50 |
+
|
| 51 |
+
| Metric | Value |
|
| 52 |
+
| --- | --- |
|
| 53 |
+
| humaneval-python | 50.0|
|
| 54 |
+
|
| 55 |
+
[Big Code Models Leaderboard](https://huggingface.co/spaces/bigcode/bigcode-models-leaderboard)
|
| 56 |
+
|
| 57 |
+
CodeLlama-34B-Python: 53.29
|
| 58 |
+
|
| 59 |
+
CodeLlama-34B-Instruct: 50.79
|
| 60 |
+
|
| 61 |
+
CodeLlama-13B-Instruct: 50.6
|
| 62 |
+
|
| 63 |
+
CodeLlama-34B: 45.11
|
| 64 |
+
|
| 65 |
+
CodeLlama-13B-Python: 42.89
|
| 66 |
+
|
| 67 |
+
CodeLlama-13B: 35.07
|
| 68 |
+
|
| 69 |
+
## lm-evaluation-harness
|
| 70 |
+
|
| 71 |
+
[Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
|
| 72 |
+
| Metric | Value |
|
| 73 |
+
| --- | --- |
|
| 74 |
+
| ARC |59.64 |
|
| 75 |
+
| HellaSwag |82.25 |
|
| 76 |
+
| MMLU | 61.33 |
|
| 77 |
+
| TruthfulQA | 48.45 |
|
| 78 |
+
| Average | 62.92 |
|
| 79 |
+
|
| 80 |
+
## Parameters
|
| 81 |
+
|
| 82 |
| | |
|
| 83 |
|------ | ------ |
|
| 84 |
| lr | 2e-4 |
|
|
|
|
| 113 |
| eeval_runtime | 0:00:25.04 |
|
| 114 |
| eeval_samples_per_second | 7.985 |
|
| 115 |
| eeval_steps_per_second | |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|